Hydroxyurea exerts bi-modal dose-dependent effects on erythropoiesis in human cultured erythroid cells via distinct pathways

Authors

  • Min Wang,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
    • *

      Present address: Institute of Hematology, Chinese Academy of Medical Sciences and Beking Union of Medical College, Tianjin 300020, P. R. China.

  • Delia C. Tang,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Wenli Liu,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Kyung Chin,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Jianqion G. Zhu,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Eitan Fibach,

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Griffin P. Rodgers

    1. The Molecular and Clinical Hematology Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author

Griffin P. Rodgers, Molecular and Clinical Hematology Branch, National Institute of Digestive and Diabetes and Kidney Diseases, National Institutes of Health, Bldg. 10, Room 9N119, 10 Center Drive, Bethesda, MD 20892, USA. E-mail: gprod@helix.nih.gov

Abstract

Summary. Hydroxyurea (HU) has been shown to increase the proportion of fetal haemoglobin (HbF) in most sickle cell patients. A low-dosage regimen increased total haemoglobin (Hb) levels in some thalassaemia intermedia patients by preferentially increasing β-globin biosynthesis. To further characterize these apparent dose-dependent effects of HU, we examined erythroid cells exposed to HU (5–100 µmol/l) in two-phase liquid culture. Low doses (from 5 to 25 µmol/l) increased Hb levels by up to 2·7-fold, and a high dose (100 µmol/l) increased Hb levels when added at d 3–6 of phase II, with no significant changes in response to HU during the late stage of phase II culture (≥ 9 d). HU exposure during d 0–3 of phase II culture increased the number of erythroid colonies to a maximum of fivefold at 5 µmol/l HU. GATA-1 mRNA was downregulated at a high dose and GATA-2 was dose dependently upregulated over a lower dosage range. Treatment with 100 µmol/l HU dramatically upregulated the death receptor DR-5, caspase 3, as determined by cDNA microarray analysis. In contrast, 10 µmol/l HU modestly upregulated mRNA levels of the early growth response gene. Our results suggest that HU exerts concentration-dependent effects on HbF production and erythropoiesis and that these two effects are mediated by distinct molecular mechanisms.

Fetal haemoglobin (HbF) can inhibit the polymerization of sickle haemoglobin (HbS) and can thereby ameliorate the clinical symptoms associated with sickle cell disease (SCD). Hydroxyurea (HU) has been used clinically to increase HbF levels (Rodgers et al, 1990; Charache et al, 1995). HU also appears to be beneficial in the treatment of some patients with β-thalassaemia (β-thal), as it reduces the imbalance in α- to non-α-globin chains (Zeng et al, 1995). Various in vitro cell culture systems have also been used to better characterize the molecular mechanisms by which HU mediates these effects. During HU treatment periods of longer than 9 weeks, the number of cultured early erythroid committed progenitors (erythroid burst-forming units, BFU-E) gradually decreased, whereas HbF levels gradually increased (Yang et al, 1997). When BFU-E derived from these patients were exposed to various concentrations of HU in vitro, a strong inverse linear relationship between the numbers of BFU-E-derived colonies and HbF levels was established (Yang et al, 1997). In a previous study, we have shown that HU, when administered at a low doses (≤10 mg/kg/d × 4 d/week), increased total Hb levels by preferentially increasing β-globin biosynthesis in some patients with thalassaemia intermedia (Zeng et al, 1995). There were no substantial changes in the biosynthesis of Gγ- and Αγ-globins over the course of 1 year in these patients. There was also an improvement in the ratio of α- to non-α-globin chains when β-thal/HbE patients were treated with HU at doses of 10–20 mg/kg/d (Fucharoen et al, 1996). Taken together, these results indicate that HU may have opposing effects on erythroid cell growth and haemoglobinization when administrated at different dosages.

A two-phase liquid culture in vitro system can be used to mimic the in vivo haematological changes that are observed in patients treated with HU. In phase I, peripheral blood mononuclear cells are first cultured in the presence of a combination of growth factors, but in the absence of erythropoietin (Epo). During this stage, early erythroid committed progenitors (BFU-E) proliferate and differentiate into erythroid colony-forming unit (CFU-E)-like progenitors. In phase II, the latter cells are then cultured in an EPO-supplemented medium, in which the CFU-E-like progenitors continue to proliferate and mature into orthochromatic normablasts and then enucleated erythrocytes. This system yields large, pure (95–98%) populations of erythroid cells that are synchronous in maturity (Fibach et al, 1993). Although these cultures have been known to contain small portion of lymphocytes, our previous data conclude that lymphocytes are dispensable to in vitro erythropoiesis (Pope et al, 2000). Cells exposed to high concentrations of HU (50–200 µmol/l) during the second part of phase II (d 6–14) were found to have an increased proportion of HbF (%HbF), increased levels of total Hb content per cell (MCH), and increased cell size (MCV), but the numbers of cells and the total amount of Hb produced were decreased (Fibach et al, 1989, 1991).

In the present study, we extended these experiments. Erythroid cells in phase II were exposed to HU either continuously (from d 0 to d 12), or pulse treated (for 3 d) over a range of HU concentrations (5–100 µmol/l). Our results showed that both continuous and pulse treatments with HU caused a significant time- and dose-dependent increase in percentage HbF. A bi-modal dose-dependent effect on cell number and total Hb was observed; that is, cell number and total Hb were increased at low concentrations of HU and decreased at high concentrations of HU. These effects were most prominent in response to pulse treatment, as compared with continuous treatment. A pulse with low concentrations of HU also increased the number of cells capable of developing erythroid colonies and downregulated the expression of the GATA-1 transcription factor. These bi-modal effects were mediated by different intracellular signalling pathways. These results provide the first evidence that HU, a well known myelosuppressive agent, can stimulate erythropoiesis and increase Hb production.

Materials and methods

Two-phase liquid culture of human adult erythroid cells.  After informed consent was obtained, mononuclear cells were isolated from the peripheral blood of healthy donors by centrifugation over Ficoll–Hypaque (1·077 g/ml; Organon Teknika Co, Durham, NC, USA). The cells were cultured as described previously (Fibach et al, 1991, 1998). Briefly, the cells were first cultured in alpha-minimal essential medium (Sigma, St. Louis, MO, USA) containing 10% preselected fetal bovine serum (FBS) (Introgen, Houston, TX, USA), 1 µg/ml of cyclosporine A (Sandoz, Basel, Switzerland) and 10% conditioned medium collected from the culture of human bladder carcinoma 5637 cell line (ATCC, Manassas, VA, USA). After 5–7 d, non-adherent cells were harvested, washed and resuspended in phase II medium composed of alpha medium, 30% FBS, 1% deionized bovine serum albumin (BSA, Sigma) supplemented with 10−6 mol/l dexamethasone (Elkin-sinn, Cherry Hill, NJ, USA), 10−5 mol/l β-mercaptoethanol (Sigma), 0·3 mg/ml of human holo-transferrin (Sigma) and 1 U/ml of human recombinant (rH) Epo (Ortho Pharmaceuticals, Raritan, NJ, USA). After exposure to Epo during 12 d of culture in phase II medium, the erythroid progenitors proliferated and matured into Hb-containing erythroid cells.

HU treatment.  Various concentrations of HU (Sigma) were added to the cultures on different days of phase II. HU remained in the cultures either continuously for 12 d or for 3 days (pulse). In pulse experiments, the cells were exposed to HU for 3 days, collected, washed twice with PBS, and re-cultured in fresh phase II medium in the absence of HU. On d 12, all cultures were terminated, and the cells were harvested and analysed.

High-performance liquid chromatography (HPLC).  Cultures were harvested and washed twice with PBS. Cells were lysed by vortexing and adding distilled water and left on ice for 5–10 min. After centrifuging for 1 min at 10 000 r.p.m, the supernatant was collected and either analysed immediately or stored at −20°C. The haemoglobins were separated and quantified by cation-exchange HPLC using a Synchropak CM 300 column (250 × 4·6 mm, Micra Scientific, Darien, IL, USA) on the GBC separation HPLC system (GBC Separations, Acton, MA, USA), as described previously (Fibach, 1998). Standard Hb solutions were used for reference (Isolab, Akron, OH, USA). Total Hb content was determined by integrating the area under the peaks. Total Hb concentration was calculated based on readings of Hb solutions of known concentrations (Fibach, 1998).

Colony formation assay.  In phase II, cells were incubated from d 0 to d 3 with different concentrations of HU. Cells were then harvested, washed twice with PBS and cloned at 1 × 105/ml in HU-free semi-solid Epo-containing medium 0·9% methylcellulose (StemCell Technologies, Vancouver, Canada), 30% FBS, 1% deionized BSA, 10−4 mol/l β-mercaptoethanol, 2 mmol/l l-glutamine, 50 ng/ml of recombinant human (rH) stem cell factor, 10 ng/ml of rH granulocyte-macrophage colony-stimulating factor (GM-CSF), 10 ng/ml of rH interleukin (IL) 3 and 3 u/ml of rH Epo. The cultures were incubated in a fully humidified incubator at 37°C in the presence of 5% CO2. Red colonies were scored on d 12. The erythroid nature of the colonies was confirmed by benzidine staining. Data are expressed as the mean ± SD from three separate experiments.

Benzidine staining.  Staining with a mixture of benzidine di-hydrochloride and hydrogen peroxide was used to identify Hb-containing cells, as described previously (Fibach, 1998). After staining, cells were scored for Hb content.

RNA isolation and reverse transcription polymerase chain reaction (RT-PCR).  RNA was isolated with TRIreagent (Molecular Research Center, Cincinnati, OH, USA) according to the manufacturer's recommended protocol. RT-PCR was carried out using murine leukaemia virus (MuLV) reverse transcriptase and Taq DNA polymerase (Applied Biosystems, Foster City, CA, USA) according to the manufacture's instructions. PCR primer pairs were as follows: γ-globin SN: 5′-AAGATGCTGGAGGAGAAAC-3′, ASN: 5′-TGCTTGCAGAATAAAGCC-3′; β-globin SN: 5′-AGGTTCTTTGAGTCCTTTG-3′, ASN: 5′-AGCCACCACTTTCTGATAG-3′; GATA-1 SN: 5′-TGATTGTCAGTAAACGGGC-3′, ASN 5′-CGTTTCTTTTTCCCTTTTCC-3′; GATA-2 SN: 5′-GGAGAACAACAAGGACAAC-3′, ASN: 5′-AAGTAAGGGGACACAGTCAC-3′, Glycophorin C (GlyC) SN: 5′-TGGGGAAGCAAGGAAATAAG-3′; ASN: 5′-AGCTGGATCAATGGCAGAC-3′. RT-PCR validation for microarray results were carried out according to the manufacturer's recommended protocol with gene-specific primers (Superarray, Bethesda, MD, USA).

Microarray hybridization.  AC133+ cells (Poietic Technologies, Gaithersburg MD, USA), were treated with 10 µmol/l and 100 µmol/l HU for 3 days respectively. Filter membranes, on which various human cDNAs sequences were microarrayed (Human GE Arrays), were purchased from Superarray. The arrays used included Human Apoptosis, Human Cell Cycle, Human p53 and Human PathwayFinder. For hybridization probes, 5 µg of total RNA from cells treated with either 10 µmol/l or 100 µmol/l HU or from untreated control cells was converted to cDNA by random hexamer labelling with Moloney MuLV (MMLV) reverse transcriptase and α-[33P]-dCTP (370 MBq/ml, NEN Life Sciences, Boston, MA, USA), according to the protocol provided by the supplier. The filter arrays were scanned in a Fuji BAS-1500 Phosphoimager (Fuji, Stamford, CT, USA). Before quantitative analysis, hybridization signals for each probe were normalized using GAPDH and β-actin cDNAs as reference controls.

Results

Pulse and continuous exposure to various concentrations of HU

Mononuclear cells derived from normal donors were subjected to the phase II culture protocol to differentiate them into erythroid cells. Cells were exposed to various concentrations of HU either continuously (from d 0 to d 12), or in a pulse treatment (exposed to HU for only the first 3 days). We found that the percentage HbF, the total level of Hb, and the number of Hb-containing cells were all regulated by HU in a dose- and time-dependent manner, as shown in Fig 1. The percentage HbF increased in a linear HU dose-dependent fashion: from 1·9% in untreated cells to 4·7% in cells pulse-treated with 100 µmol/l HU and from 1·9% to 5·4% in cultures treated continuously with 100 µmol/l HU (Fig 1A). HU showed a bi-modal dose–effect on the number of Hb-containing cells and on total Hb levels. Continuous exposure to high concentrations of HU (100 µmol/l) dramatically decreased both the number of Hb-containing cells and the Hb level by approximately 90% (Fig 1B and C). Similarly, pulse exposure to 100 µmol/l decreased each of these parameters by approximately 30% (Fig 1B and C). In contrast, at low concentration (25 µmol/l), HU augmented both parameters. The stimulatory effect was more pronounced in pulse-treated cultures (where the number of Hb-containing cells and the Hb level were increased by 1·5- and 1·9-fold respectively) than in continuously treated cultures, which showed only a modest response to 25 µmol/l HU (Fig 1B and C). The stimulatory effect of pulse treatment with low doses of HU on cell number and Hb content was observed with doses of HU as low as 5 µmol/l (Fig 2B and C). These results implied that pulse and continuous treatment with HU have different effects on cell differentiation and proliferation, although both pulse and continuous exposures to HU dose-dependently increased percentage HbF (Figs 1A and 2A).

Figure 1.

Effect of pulse versus continuous exposure to hydroxyurea (HU) on various haematological parameters in cultured erythroid cells. HU, at the indicated concentrations, was added to erythroid cells on d 0 of phase II for either 3 d (pulse) or 12 d (continuous), as described in Materials and Methods. On d 12, the cells were harvested and analysed for percentage HbF (A), total Hb (B) and number of benzidine-positive cells (C).

Figure 2.

Effect of HU pulse treatment at different times during the culture. Phase II cultures were pulse treated (for 3 d) with the indicated concentrations of HU on different days (d 0–3, d 3–6, d 6–9 or d 9–12), as indicated. The cells were harvested on d 12 and analysed for percentage HbF (A), total Hb (B) and number of benzidine positive cells (C).

Pulse exposure at different stages of erythroid cells

To study the effect of HU on cells at different stages of maturation, cultures were pulse exposed (for 3 d) to low concentrations of HU on different days of Phase II. Consistent with previous observations, HU induced a dose-dependent increase in percentage HbF. As shown in Fig 2A, the percentage HbF was increased when HU was added at any time during Phase II. However, HU had the greatest effect on percentage HbF when cells were pulse-treated during the first 6 d. The percentage HbF increased from 2·0% to 4·7% when cells were pulse-treated with 100 µmol/l HU on d 0–3 and from 1·9% to 5·5% when cells were pulse-treated with 100 µmol/l HU on d 3–6 (Fig 2A). Although modest increases in percentage HbF were observed when HU was added after d 6, these effects did not achieve statistical significance. The bi-modal effect of HU on cell number and on total Hb was most pronounced in cultures pulse-treated during the first 9 d (Fig 2B and C). Exposure to HU during the late stage of phase II (at d 9–12) did not have a significant effect on the number of Hb-containing cells or on total Hb levels (Fig 2B and C). Taken together, the data indicate that cells must be exposed to HU during the early or intermediate stages of maturation to respond with alterations in erythroid cell growth and haemoglobinization.

Pulse exposure to HU on erythroid colony formation

To evaluate the effect of HU on erythropoiesis, we analysed the colony-forming potential of cells. The cloning potential increased from 47 per 105 cells in the absence of HU to a maximum of 236 per 105 cells at 5 µmol/l HU (Fig 3). At higher doses, ranging from 10 to 100 µmol/l HU, the cloning potential progressively decreased to near control levels. These results further confirmed the bi-modal dose–response effects of HU on erythroid cell growth.

Figure 3.

Effect of pulse exposure to HU on the cloning potential. Phase II cultures were pulse treated (from d 0 to d 3) with the indicated concentrations of HU. The cells were then harvested, washed, counted and cloned in HU-free, Epo-containing semi-solid medium. Erythroid colonies were scored on d 10–12.

Effects of low and high concentration of HU on globin gene expression

Next, we determined globin gene expression levels by RT-PCR. Cells were pulse-treated on d 0–3 with HU at doses ranging from 0 to 100 µmol/l. We found that β-globin mRNA levels were higher in cultures pulse-treated with low concentrations of HU (5 and 10 µmol/l). On the other hand, γ-globin mRNA was increased only at high concentrations of HU (100 µmol/l) (Fig 4A). These findings suggest that HU-mediated regulation of globin mRNA is consistent with protein expression.

Figure 4.

Effect of HU on expression of α-globin, β-globin and GATA transcription factor mRNA levels. Phase II cultures were pulse treated (from d 0 to d 3) with the indicated concentrations of HU. The cells were harvested at the end of the culture. Total RNA was isolated and analysed by RT-PCR for (A) γ-globin and β-globin mRNA levels and (B) GATA-1 and GATA-2 mRNA levels.

Effects of HU on GATA-1 and GATA-2 expression

To examine whether the GATA-1 and GATA-2 expression levels are regulated by different concentrations of HU, RT-PCR was performed on erythroid cells after a 3-d exposure to HU at the beginning of phase II. At low concentrations (5–50 µmol/l), HU did not have any significant effect on GATA-1 mRNA levels, whereas GATA-1 mRNA was decreased by more than 50% in cells pulse-treated with 100 µmol/l (Fig 4B). Thus, at a high dose (100 µmol/l), HU downregulated the expression of GATA-1 mRNA. In contrast, HU progressively induced a dose-dependent increase in GATA-2 mRNA, to a maximal level of fivefold over baseline levels at 50 to 100 µmol/l HU. The differential effects of HU on expression of the GATA-1 and GATA-2 transcription factors suggests that low versus high doses of HU might regulate distinct signal transduction pathways.

Effect of low and high concentrations of HU on gene expression profiling

We used cDNA microarrays as an initial step towards further characterizing the mechanisms of differential effects of low and high doses of HU on gene expression of γ and β globin mRNA levels. Specific microarrays that included genes involved in cell cycle control (including cdk, cyclin, c-myc, gadd45, mdm2, p21, and p53), apoptosis (including bad, bax, bcl-2, bcl-w, bcl-x, caspase, fas, gadd45, mdm2 NFkB, Rb, DR5, trail), and various signal transduction pathways were used to evaluate the effect of HU on gene expression patterns on a large scale. The signal pathways are mitogenic, stress, NFκB, NFAT, anti-proliferation/TGFβ, Wnt, p53 and CREB signal pathways (including ATF-2, CD5, c-fos, c-myc, CYP19, egr-1, Fas, hsp90, IL-2, iNOS, p57, pig7). As the primary cultures grown in the two-phase system contain a relatively low percentage of non-erythroid cells, we used cultured AC133+ human bone marrow stem cells to characterize the gene expression profile in response to HU treatment. AC133+ cells are early progenitor, which are exclusively CD34+ cells (Miraglia et al, 1997; Yin et al, 1997). These cells grew readily in conditions similar to the two-phase liquid system that we used in the experiments described above and, with Epo induction, resulted in synchronized erythroid cells. However, we limited the CD34+ use to the gene profiling experiments only and not the pharmacokinetic studies detailed above given the comparative cost of AC133+ cells versus the relative abundance of buffy coat cells. In this experiment, AC133+ cells were exposed to low (10 µmol/l) and high (100 µmol/l) doses of HU for three days.

There were substantial differences in the expression of several groups of genes in AC133+ cells treated with 0, 10 and 100 µmol/l HU, summarized in Table I. Treatment with 100 µmol/l HU dramatically induced expression of death receptor-5 (DR5) mRNA 80-fold over control levels (Fig 5A and B, position A8, B8) and increased the expression of caspase 3 (Fig 5A and B, position C3, D3). This dose of HU also significantly upregulated the expression levels of caspase 10, the pro-apoptotic gene bax, and the antiapoptotic gene bcl-xL.

Table I.  Comparison of induced gene expression between 10 µmol/l and 100 µmol/l hydroxyurea (HU)-treated AC133+ cells using pathway-specific gene expression microarray hybridization.
GeneDescription of geneGenBank accession no.Pathway-specific array*Average increase (fold)**
10 µmol/l HU100 µmol/l HU
  • *For complete lists of all genes included in each array, see http://www.superarray.com

  • **

    Average of increased expression compared with untreated, signal intensity normalized with reference genes (at least three experiments).

  • N, no difference in gene expression from untreated AEC cells; Apo, apoptosis GEArray; Ccy, Cellcycle GEArray; p53, p53 GEArray; PF: PathwayFinder GEArray.

baxBCL2-associated X proteinL22474ApoN7·5
bcl-x LBCL2-like 1Z23115ApoN11·5
DR5Homo sapiens TRAIL receptor 2AF016266ApoN80·0
Caspase 10Apoptosis-related cysteine proteaseU60519ApoN3·7
Caspase 3(cpp32)Caspase 3, apoptosis-related cysteine proteaseU13737ApoN2·3
gadd45DNA-damage-inducible transcript 1M60974Apo, Ccy, p53, PFN10·8
mdm2Human homologue of mouse double minute2;Z12020Apo, Ccy, p53, PFN3·5
p53-binding protein    
P21Wafl(P21Cip1)Cyclin-dependent kinase inhibitor 1AL47233Apo, Ccy, p53, PFN18·7
(p21, Cip1)    
skp1Cyclin A/CDK2-associated p19(Skp1)U33760CcyN2·5
GPX1Glutathione peroxidase 1M21304p53N3·6
PCNAProliferating cell nuclear antigenJ04718p53N2·0
egr-1Early growth response 1X52541PF2·0N
hsp90Heat-shock protein, 90 kDaX15183PF2·32·6
Figure 5.

Gene expression profiling of cells treated with high and low concentration of HU. RNA isolated from AC133+ cell treated with no HU (A) and 100 µmol/l HU (B) were labelled in first strand cDNA synthesis and hybridized to the apoptosis array membrane (see company's web site for details of the gene spotted - www.superarray.com). RT-PCR was used to validate the results from array of cells treated with 0, 10 and 100 µmol/l (C).

Interestingly, we also found that mRNA levels of several cell cycle control genes were increased in response to 100 µmol/l HU, including p21cip, skp-1, gadd45 and mdm2. On the other hand, exposure to a low dose of HU (10 µmol/l) increased expression of the immediate early response gene erg-1. RT-PCR with gene-specific primers was carried out to validate the microarray results, and the results of representative experiments are shown in Fig 5C.

These data suggest that treatments with low and high concentrations of HU regulate different genes through separate signalling pathways. This differential regulation could be the basis for the divergent effects of low versus high doses of HU that we observed on cell differentiation and proliferation.

Discussion

Hydroxyurea is a myelosuppressive agent that increases HbF levels and has been used clinically to treat SCD and β-thalassaemia. Our study clearly showed that both continuous and pulse treatments with HU caused a significant dose-dependent increase in percentage HbF. The magnitude of this effect varied, depending on the specific phase of the 12 d period at which the drug was added to the culture. Both the number of Hb-containing cells and total Hb level were increased by low concentrations of HU and decreased by high concentrations of HU. These effects were most pronounced in response to pulse (3 d) treatment, as compared with continuous treatment. Pulse treatment with low concentrations of HU also increased the number of cells that are capable of developing erythroid colonies.

Our results provide the first evidence that HU can paradoxically stimulate erythropoiesis and concomitantly increase the production of Hb. The failure of previous studies to show this effect in primary cultures may be related to the relatively high doses and the long exposure times used in these studies (Mankad et al, 1994; Yang et al, 1997). In our experiments, enhanced erythropoiesis was observed when phase II cultures were pulse-treated for 3 d during the early stages (the first 6 d) with low concentrations of HU (< 50 µmol/l). In vivo studies have shown that continuous infusion of HU at 20 mg/kg will maintain a serum concentration of HU at 100 µmol/l (Fabricius & Rajewsky, 1971). The exposure regimen in our study closely resembles the in vivo pharmacokinetics of HU treatment. By d 5 in phase II, the cells morphologically resembled human CFU-E retrieved from methylcellulose (Sawada et al, 1987; Miller et al, 1999). This implies that if HU were added to the culture before or at the CFU-E stage, it would be able to stimulate eythropoiesis and to increase the Hb production.

We also found striking evidence from our in vitro experiments that HU displays a bi-modal dose–effect relationship. That is, total Hb, but not of HbF, was increased in response to treatment with low concentrations of HU. In contrast, HbF, but not total Hb, was increased in response to treatment with high concentrations of HU. This in vitro data correlates closely with clinical observations (Rodgers et al, 1990; Charache et al, 1995; Zeng et al, 1995). We have shown previously that when certain patients with thalassaemia intermedia were administered a relatively low dose of HU (6–8 mg/kg/d × 4 d/week), total Hb was increased by a preferential increase in β-globin biosynthesis (Zeng et al, 1995). With these protocols, patients maintained Hb levels greater than 3 g/dl above pretreatment levels, and the HbF was not increased during the treatment period. In a study of Brazilian patients with either SCD or sickle β-thalassaemia, there was an improvement in the concentration of Hb (range: 0·7–2·0 g/dl) at a HU dose of 15 mg/kg/d, but this concentration did not increase significantly when the HU dose was raised to 20 mg/kg/d. In contrast, Hb F levels increased significantly (range: 1·0–18·1%) on the 15 mg/kg/d dose, and continued to increase on the higher dose of 20 mg/kg/d (Lima et al, 1997). These in vivo data support the in vitro results reported herein, confirming that HU exerts concentration-dependent effects on the production of HbF and on the level of effective erythropoiesis, and that these two effects can be dissociated.

The GATA protein transcription factors play key roles in controlling the proliferation and differentiation of haematopoietic cells. We hypothesized that HU may affect the expression or alter the function of haematopoietic-related transcription factors, such as GATA-1 and GATA-2, to participate in the processes of proliferation and differentiation of erythroid cells. GATA-1 is expressed at higher levels in mature erythroid cells, mast cells and megakaryocytes, and at relatively lower levels in multipotent progenitor cells, which are known to express differentiation-specific genes and to suppress the expression of genes involved in proliferation of committed erythroid progenitor (Orkin, 1992; Weiss & Orkin, 1995a; Ohyashiki et al, 1996). GATA-1 facilitates the survival and maturation of erythroid precursors by preventing apoptosis (Weiss & Orkin, 1995b). In contrast, GATA-2 is preferentially expressed in more primitive cells, such as haematopoietic stem cells and multipotential progenitor cells (Briegel et al, 1996). Overexpression of GATA-2 blocks normal haematopoiesis (Persons et al, 1999). Increased GATA-2 levels have been correlated with increased γ-globin gene expression in K562 cells (Ikonomi et al, 2000). The differential effects of HU on GATA-1 and GATA-2 may explain the increases in erythroid colonies, Hb-containing cells, and total Hb production in response to low concentrations of HU and the increase in the percentage HbF in response to high concentrations of HU. Our results revealed that GATA-1 expression was stable at a relatively low HU concentration and that GATA-1 expression was moderately decreased when erythroid cells were exposed to 100 µmol/l HU. GATA-2 expression was relatively low in control cells and in cells exposed to low concentrations of HU. However, GATA-2 mRNA was progressively increased with increasing doses of HU. The decrease in GATA-1 mRNA at high concentrations of HU would be predicted to delay the maturation of definitive proerythroblasts into mature erythrocytes and thereby lead to the accumulation of HbF-producing cells. On the other hand the increased GATA-2 expression in response to high doses of HU correlated with the increases observed in gamma globin gene expression. The ratio of GATA-2 to GATA-1 mRNA may also play a role in the bi-modal effects of HU. At a ratio of 4:1 or less, as was observed at low concentrations of HU, the erythroid cells continue terminal differentiation and proliferation. When the ratio reached a threshold greater than 4:1, as was observed at high concentrations of HU, erythropoiesis was delayed and HbF levels were increased.

As determined by cDNA microarray analysis, exposure to a high concentration of HU (100 µmol/l) strongly induced the expression of a number of apoptosis related genes (Table I), most notably the death receptor, DR-5. Death receptors are members of a family of surface molecules that trigger activation of caspases and thereby induce apoptosis in various cell types. Interestingly, there is evidence suggesting that the death receptor ligand induces caspase-mediated degradation of GATA-1 and negatively regulates erythropoiesis (De Maria et al, 1999). Caspase 3 expression was also increased in response to high concentrations of HU. Taken together, it is possible that HU, acting as a ligand for the death receptor, triggers the caspase pathway and degrades GATA-1, thereby negatively regulating erythropoiesis. The inhibition of erythropoiesis will lead to increased production of HbF. Our data provides new insights into the possible mechanism by which HU increases HbF levels.

GATA-1 and Epo have been reported to cooperatively induce bcl-xL expression and to promote erythroid cell survival (Gregory et al, 1999). Our findings demonstrate that expression of GATA-1 was slightly decreased, GATA-2 was significantly increased, and bcl-xL was increased at high concentration of HU. It is possible that increased expression of GATA-2 may complement the effects of GATA-1 on inducing bcl-xL expression. It has been reported that activation of the Epo receptor increased the ratio of GATA-1:GATA-2 and concomitantly downregulated bcl-xL via a protein kinase C-dependent pathway (Tsushima et al, 1997). In our culture system, HU acted in the opposite manner. That is, the GATA-1:GATA-2 ratio was decreased and bcl-xL mRNA was increased. Increased bcl-xL promotes the survival of erythroid cells and, therefore, protects the cells from cytotoxic effects of HU and apoptosis. The cells retain their ability to produce HbF.

Treatment with a low dose of HU (10 µmol/l) caused a twofold induction of egr-1 mRNA expression. Egr-1, also known as TIS8, krox-24, and NGFI-A, is a zinc finger transcription factor that is induced in various cell types in response to diverse stimuli. Egr-1 expression had been related to the expression of several important growth factors and signalling molecules, such as platelet-derived growth factor (PDGF), fibroblast growth factor (FGF-2), the extracellular signal-regulated kinase (ERK), and c-jun NH2-terminal kinase (JNK) pathways (Silverman & Collins, 1999). Egr-1 has been shown to be important for regulating macrophage differentiation in response to viral infections and inflammatory states (Henderson et al, 1994). Egr-1 is also induced in response to hypoxia (Yan et al, 1998) and is increased in compensatory lung growth. (Landesberg et al, 2001). Recently, the egr-1 gene was implicated as a potential radiation response gene marker in prostate cancer (Ahmed et al, 2001). The increase in egr-1 gene expression suggested that, at low concentrations, HU might function like a growth factor to upregulate egr-1 gene expression and to promote the maturation of erythroid progenitors by increasing the total haemoglobin level.

In summary, we have identified a bi-modal dose–effect of HU on cultured human erythroid cells. Taken together with its in vivo effects, our in vitro results suggest that HU exerts concentration-dependent effects on HbF production and erythropoiesis. These two effects can be dissociated and appear to be mediated by different molecular mechanisms. Given that low-dose HU increases both the proliferation and differentiation of erythroid progenitors, we are currently examining whether these in vitro effects can be replicated in progenitors derived from patients with refractory anaemia with and without ring sideroblasts (Preisler et al, 1999).

Acknowledgments

We are grateful to Mr. David Lewis for his technical assistance. The first two authors contributed equally to this work.

Ancillary