βMinor-globin messenger RNA accumulation in reticulocytes governs improved erythropoiesis in β thalassemic mice after erythropoietin complementary DNA electrotransfer in muscles

Efficient reactivation of fetal hemoglobin (HbF) in adults would be a convenient approach to treat hemoglobinopathies secondary to β-globin defects. Compounds such as 5-azacytidine, hydroxyurea, butyric acid, or derivatives have been intensively investigated to effect that goal.1 So far, clinical benefits appear limited, at least in β thalassemias.1-4 

Increased fetal erythropoiesis in response to intense stimulation by erythropoietin (Epo) was previously observed in animal models and in erythroid cell cultures. Severe anemia stimulates β^C-globin expression in sheep.5Injections of high doses of recombinant human Epo (rhuEpo) increased βminor-globin (βmin-globin) chain synthesis in mice6and the percentage of circulating fetal cells in normal or anemic baboons.7,8 We showed that continuous secretion of high quantities of murine Epo from genetically modified muscles increased βmin/α-globin chain synthesis ratio in β thalassemic mice and improved the diseased phenotype.9 In humans, a high concentration of rhuEpo increased the proportion of colonies containing HbF in erythroid cell cultures.10 However, trials with rhuEpo in patients with β thalassemia have been relatively disappointing thus far, although clinical improvements were occasionally observed.4,11-15 A current widely accepted view is that reactivation of fetal erythropoiesis requires a more intense stimulation in humans than in the animal models that have been explored. A low threshold may account for clinical responsiveness in certain individuals, including transfusion-dependent β thalassemic patients.15 Encouraging results were also obtained in a trial combining high doses of Epo with hydroxyurea in patients with β thalassemia intermedia.4 Tolerance was good in all reported trials. Major drawbacks impairing further assessment of this treatment are cost effectiveness and the theoretical risk of worsening of bone marrow expansion and bone disease.

Molecular mechanisms governing a reactivation of fetal erythropoiesis are imperfectly understood. Smith and colleagues recently reported that hydroxyurea, 5-azacytidine, or butyric acid modulates the levels of both γ- and β-globin messenger RNA (mRNA) in human adult erythroid cells.16 Increased β^C-globin mRNA amounts were also observed in anemic sheep5 or in sheep bone marrow cultures stimulated with high doses of Epo.17It has been proposed that a more efficient translation of βmin-globin mRNA than α-globin mRNA would account, at least partly, for the compensatory increase of βmin-globin synthesis observed in β thalassemic mice, in which serum Epo concentration is elevated.18 However, how Epo improves the phenotype of β thalassemic erythrocytes has not been directly investigated so far. The increased ratio of newly synthesized βmin/α-globin chains that has been observed in β thalassemic mice receiving or secreting high doses of Epo may theoretically result from various, possibly combined mechanisms acting on βmin-globin gene transcription, βmin-globin or α-globin mRNA stability, mRNA translation efficiencies, and globin chain proteolysis. Moreover, additional indirect effects of Epo that would participate in the appearance of an effective erythropoiesis cannot be excluded as possible actions on membrane defects, oxidative processes, or iron compartmentalization, because each of these phenomena plays a crucial role in pathophysiology.19 

We investigated these issues in the C57Bl/6^Hbbth mouse model of β thalassemia.20 These animals have a complete deletion of the mouse βmajor-globin gene. Because of the compensatory elevation of βmin-globin chains, which decreases the amount of unpaired α chains, the mouse phenotype is more relevant to that of human β thalassemia intermedia than to β thalassemia major.21 The stimulation of βmin-globin expression in C57Bl/6^Hbbth mice is considered as a suitable model of the reactivation of fetal erythropoiesis in humans.22-25 We have induced Epo oversecretion in these animals by introducing an expression vector for murine Epo into skeletal muscles. Gene transfer was performed by the intramuscular injection of naked DNA associated with an electric shock. In comparison with naked DNA injection alone, this method increases gene expression level and persistence.26 Various parameters were measured on blood samples taken from these animals before and twice after treatment. Values were compared in individual animals using statistical analysis. Investigations included a survey of globin mRNA content in reticulocytes, as measured by real-time quantitative polymerase chain reaction (PCR), a study of globin chain synthesis, and the analysis of membrane proteins components and iron compartments. Data indicate that the accumulation of βmin-globin mRNA accounts for the appearance of an effective erythropoiesis in response to Epo overproduction in β thalassemic mice.

Plasmid MCK3.3/rtTA2^S-S2 contains the coding sequence for rtTA2^S-S2,27 a variant of the original tetracycline-inducible transactivator rtTA.28Expression is controlled by the muscle creatine kinase (MCK) 3.3-kilobase (kb) promoter29 and a polyadenylation signal derived from the bovine growth hormone gene.30 The murine Epo complementary DNA (cDNA) has been inserted in the same construct downstream of heptamerized tetO sequences associated with a minimal human cytomegalovirus promoter.31Plasmid DNA was prepared by standard double CsCl gradient purification and resuspended in sterile saline solution. Avertine-anesthetized C57Bl/6^Hbbth mice, 3 to 4 months old (bred from animals obtained at the Jackson Laboratory, Bar Harbor, ME), were injected in the quadriceps muscles with 100 μg plasmid DNA. Electrostimulation was performed immediately after injection, as described previously.26 Electrical shocks consisted of 8 trains of 10³ pulses (45 V, 50 mA, 200 μs/phase). The electric field was applied through a pulsar 6 bp-a/s stimulator (FHC, Bowdoinham, ME). Injected muscles were resected at sacrifice (22 weeks) and high-molecular-weight DNA was prepared. Vector DNA detection by PCR showed persistence in all treated animals. Doxycycline-HCl (Sigma, Saint-Quentin Fallavier, France) was dissolved in the drinking water to a final concentration of 200 μg/mL with 5% sucrose.

Hemoglobin (Hb), hematocrit (Hct), and red blood cell (RBC) counts were determined by standard procedures. Reticulocytes were counted after staining with brilliant cresyl blue. Serum Epo concentrations were measured by enzyme-linked immunosorbent assay (ELISA).32 The assay relies on the capture of mouse Epo on microtitration plates coated with an antihuman Epo monoclonal antibody (mAb; R&D Systems, Minneapolis, MN) and revelation with a horseradish-conjugated antihuman Epo mAb (Roche Molecular Biochemicals, Mannheim, Germany). Quantification with recombinant mouse Epo indicates a sensitivity of 1 mU/mL.

The method is based on the detection of fluorescence generated by the TaqMan probe degradation.33 The probe anneals between 2 amplification primers and is degraded by the nucleolytic activity of the ampliTaq Gold DNA polymerase at each polymerization step. Fluorescence is monitored in real time. Intensity is related to the initial number of DNA copies, which can be assessed be determining the threshold cycle (Ct).34 

Total RNA extraction from blood samples and first-strand synthesis were performed using standard procedures (RNeasy mini-kits, Qiagen, Santa Clarita, CA; oligodeoxythymidine priming, Clontech, Palo Alto, CA). Quantitative real-time PCR of cDNA was carried out with specific double fluorescently labeled probes in an Abi Prism 7700 Sequence Detector (Perkin-Elmer Applied Biosystems, Norwalk, CT). 6-Carboxyfluorescein (FAM) was the 5′ fluorescent reporter and tetramethylrhodamine (TAMRA) the 3′ end quencher. Probes were designed to span exon junctions to prevent amplification of contaminating genomic DNA. The following primers and probes were used: βminor-globin forward primer: 5′-ACCTGGGCAAGGATTTCACC-3′; βminor-globin reverse primer: 5′-CCACTCCAGCCACCACCT-3′; βminor-globin probe: 5′-FAM-TGCTGCACAGGCTGCCTTCCAG-TAMRA-3′; α-globin forward primer: 5′-AATATGGAGCTGAAGCCCTGG-3′; α-globin reverse primer: 5′-AACATCAAAGTGAGGGAAGTAGGTCT-3′; α-globin probe: 5′-FAM-AAGGATGTTTGCCAGCTTCCCCACTACT-TAMRA-3′.

Amplification reactions (25 μL) contained 10 μL sample cDNA or standard DNA, 1 × TaqMan buffer, 5 mM MgCl₂, 0.2 mM dA/dC/dGTP, 0.4 mM dUTP, 0.125 U AmpliTaq Gold, 0.2 mM primers (forward and reverse), and 0.1 mM TaqMan probe. Amplification was performed by 50 cycles of 95°C, 15 seconds and 60°C, 1 minute. Standard amplification curves were obtained by serial dilutions of the cDNA of interest, which concentration was accurately determined.16GAPDH mRNAs were used as internal controls for extraction, reverse transcription, and amplification. Data were edited using the Primer Express software (Perkin-Elmer, Applied Biosystems, Foster City, CA).

Globin chain synthesis was analyzed by metabolic labeling as previously described.9 Briefly, blood cells were washed and resuspended in methionine- and cystein-free RPMI 1640 medium for 30 minutes at 37°C before a 20-minute labeling with 250 μCi 35S/methionine/cystein (Amersham Life Science, Arlington Heights, IL) and a 1-hour chase. Protein extracts (50 μg) were analyzed on urea-Triton-polyacrylamide gel electrophoresis. Gels were dried for quantification on a Phosphorimager (Molecular Dynamics, Sunnyvale, CA). Erythrocyte ghost extracts35 and inside-out membranes (IOMs)36 were prepared as described.

Surface immunoglobulins were detected using a bead-rosette assay performed on RBC suspension.37 Ten microliters of polystyrene beads (Dynabeads M-450; Dynal, Great Neck, NY) coated with affinity-purified goat antimouse IgG were incubated with 2.5 × 10⁵ RBCs for 1 hour, shaking gently at room temperature. The percentage of RBCs with attached beads was determined using direct light microscopy. At least 400 RBCs were counted in each experiment. Addition of soluble mouse immunoglobulin to the suspension medium reduced the proportion of erythrocyte-bound immunoglobulins in untreated β thalassemic mice to that observed in normal mice, indicating that the RBC-bead interaction was immunoglobulin mediated.

Free iron (nonheme, nonferritin iron) was determined by reactivity with ferrozine within 2 minutes of incubation. Freshly prepared ghosts or IOMs were dissolved in 500 μL 0.6% sodium dodecyl sulfate (SDS) in 0.2 M sodium acetate, pH 4.5, and incubated 5 minutes at room temperature in 0.35 M ascorbic acid, 10 M sodium metabisulfide in 0.2 M sodium acetate, pH 4.5. The color developing solution (100 μL, 200 mg ferrozine and 1.25 g thiourea dissolved in 50 mL water) was added for 2 minutes and the absorbance was measured at 562 nm. Each measurement was made in duplicate. Concentration of total nonheme iron (membrane iron plus any other iron such as ferritin iron) was determined after a 24-hour incubation, after which no further reaction occurred. A millimolar extinction coefficient of 27.9 was used for determining iron concentration. Total heme iron content (including free heme, hemoglobin, and hemichrome) was measured by absorbance at 398 nm of either a known amount of membrane preparations or of IOMs dissolved in formic acid. A millimolar extinction coefficient of 83.5 ± 1.8 at 398 nm was used for determining heme iron concentration. All reagents were rendered iron-free by treatment with a chelating resin (Sigma Chemical, St Louis, MO).

Survival of RBCs was measured using a nonradioactive method.38 Mice were injected intravenously 3 times over a 24-hour period with 1 mg NHS-X-Biotin (Calbiochem, San Diego, CA). For analysis, 2.5 μL capillary blood obtained by tail puncture was washed 3 times in phosphate-buffered saline (PBS) with 0.5% bovine serum albumin (BSA) and incubated 30 minutes at 4°C with 5 μL ALEXA-conjugated streptavidin (Pierce Chemical, Rockford, IL). Samples were analyzed using a FACScan (Becton Dickinson, Mountain View, CA). Labeled cells were expressed as a percentage of total cell count (5000 per sample). Routinely, more than 94% of circulating RBCs stained positive 24 hours after the last injection.

Statistics were performed using the SPSS software (SPSS, Chicago, IL).

The effect of a transient Epo secretion from muscles was investigated in 9 C57Bl/6^Hbbth homozygous β thalassemic mice. Animals were studied regarding several parameters relevant to erythropoiesis at 3 time points: 1 week before treatment, and 4 and 11 weeks after treatment. Data collected in individual animals before and after treatment were analyzed using the nonparametric Wilcoxon 2–related samples test to search for relationships between the various explored parameters. Treatment consisted of a single intramuscular injection of DNA encoding a tetracycline-inducible expression vector for murine Epo. Electrostimulation was performed immediately after injection. We expected that inducible expression would allow choice of various levels of Epo secretion, as we previously observed,26 thus facilitating the analysis of relationships between parameters. Doxycycline was given in drinking water (200 μg/mL), starting 1 day after DNA injection. Because this single dosage induced a large range of effects depending on the individual animal, it was continuously provided until sacrifice at 22 weeks.

Animals suffered severe anemia before treatment (Hct, 32.9% ± 1.06%; Hb, 90 ± 4.7 g/L; RBC count, 9.7 ± 1.6 × 10⁹/mL) with characteristic hypochromia (mean corpuscular hemoglobin concentration [MCHC], 27.2 ± 1.5 pg/dL), anisocytosis, and microcytosis (mean corpuscular volume [MCV], 35 ± 1.2 fL) (Table 1 and Figure 1A). Serum Epo concentrations, which were below background detection level in normal mice, ranged from 18 to 61 (mean 34 ± 16) mU/mL in β thalassemic mice (Table 1 and Figure 1B).

All treated animals showed increased serum Epo concentrations 4 weeks after injection (200 ± 75 mU/mL). Individual values were distributed over a large range (132.6-325.4 mU/mL) (Table 1 and Figure1B). Concentrations decreased to 150 ± 47 mU/mL between weeks 4 and 11. They were below the detection level in sera of mice 1, 3, 5, and 8 at week 22. Thus, plasmid DNA electroinjection induced a transient Epo secretion in β thalassemic mice, with peak values variable between animals. When given to normal mice, a similar treatment plan resulted in a more stable Epo secretion.26 

The Hct, Hb concentration, and RBC counts increased in all treated mice (Table 1). Values for serum Epo concentrations were distributed over a large range. Some animals became polycythemic, but others were normocythemic and yet others remained anemic. Wilcoxon tests showed significant relationships between Hct, Hb, RBC values, and serum Epo concentrations (P = .0077 at 4 weeks andP = .0117 at 11 weeks). These data confirmed that increasing serum Epo concentration improves erythropoiesis in β thalassemic mice.6,9,39,40 

In mouse 1, which was polycythemic at 4 weeks, erythropoietic stimulation was associated with a correction of MCHC and MCV (33.4 g/dL and 41.3 fL, respectively). This mouse died of polycythemia 5 weeks after gene transfer. In mice 2 through 7, in which Hct values ranged between 45.3% and 55.6% (mice 3-7) and 70.3% (mouse 2) at 4 weeks, improvement of hypochromia and microcytosis was partial. In mice 8 and 9, in which serum Epo concentrations were only slightly increased, Hct values remained below 40% and hypochromia and microcytosis were almost unchanged. These results show that the improvement of hypochromia and microcytosis was limited unless Hct reached high values.

High reticulocyte cell counts and percentages in the blood of β thalassemic mice reflected chronic erythropoietic stimulation. A decreased proportion of circulating reticulocytes was observed in all treated mice (Table 1). This indicated that the appearance of a more effective erythropoiesis was accompanied by a reduction of erythroid cell proliferation, as previously reported.9 The reduction largely varied depending on animals (eg, from 28% to 9.6%, 30.7% to 20.5%, and 32.2% to 26% in mice 1, 5, and 8, at 4 weeks, respectively). Wilcoxon tests showed a significant relationship between reticulocyte proportions in the blood and serum Epo concentrations (P = .0077 at 4 weeks; P = .0117 at 11 weeks). However, reticulocyte counts remained elevated with respect to the increased blood mass (between 18.7 × 10⁵/μL and 29.4 × 10⁵ /μL at 4 weeks versus 3.1 ± 0.4 × 10⁵/μL in normal mice), indicating that erythropoiesis was still accelerated.

Reduced RBC survival is a hallmark of β thalassemia. We measured this parameter by labeling erythrocytes in vivo with NHS-X-biotin.37 The half-life of normal erythrocytes measured by this method was 18.9 ± 1.9 days (n = 3), a value consistent with that obtained using radiolabeling.37,38Survival of erythrocytes in untreated β thalassemic mice was 9.2 ± 1.1 days. Survival of erythrocytes was 16.9 days when measured 14 weeks after treatment in mouse 2, whose Hct value was 55.6% at that time. This result showed that improved erythropoiesis was associated with increased RBC survival.

Beneficial effects of Epo on β thalassemic erythropoiesis have been previously reported in mice,6,9,39,40 and to a lesser extent in humans.11-15 However, little is known about the mechanisms by which Epo induces the appearance of effective erythropoiesis. We took advantage of the large distribution of serum Epo values and hematologic parameters in treated animals to investigate possible relationships between the amounts of βmin-globin mRNA present in reticulocytes at different time points and various parameters relevant to the appearance of an effective erythropoiesis.

We followed globin mRNA amounts in the circulating reticulocytes of living animals by quantitative real-time PCR. Total RNA was extracted from blood samples of β thalassemic mice before treatment and at weeks 4 and 11 after DNA injection. cDNA was synthesized by reverse transcription. Primers and probes were designed for the amplification of both murine βmin- and α-globin cDNA. The mRNA copy numbers were estimated from cDNA amounts, which were quantified by the amplification of a reference DNA. These values represent an operational quantification intended to allow comparisons between time points and animals.

Equivalent α-globin mRNA levels were measured in untreated and treated mice, indicating that Epo did not affect the expression of theα-globin gene (Table 1 and Figure 1C). Estimated copy numbers ranged between 45 and 58 mRNA molecules per reticulocyte. The estimated copy number of βmin-globin mRNAs was 16.4 ± 1.6/reticulocyte in untreated mice. Values were 2- to 3-fold more abundant 4 weeks after DNA injection in mice 2 through 7 (Table 1and Figure 1D). They further decreased between weeks 4 and 11, but remained higher than before treatment in mice 2, 3, 4, and 7. A Wilcoxon test showed a strong relationship between estimated βmin-globin mRNA copy numbers and serum Epo concentrations (P = .0117 at weeks 4 and 11). Thus, Epo induced either the accumulation of βmin-globin mRNAs or the production of erythroid cells in which βmin-globin mRNAs accumulate. The βmin-globin mRNA copy numbers were also significantly related to Hct values (P = .0117 at weeks 4 and 11). They remained unchanged in mice 8 and 9, which showed little increase of serum Epo concentration and did not improve erythropoiesis.

In normal mice, the βmin plus βmaj/α-globin mRNA ratio is close to 1. Before treatment, the βmin/α-globin mRNA ratio in our β thalassemic mice was 0.31 ± 0.01. Values rose above 0.55 in mice 2 through 7 after treatment, reaching up to 0.9 in mouse 2 (Figure 1E). However, values never reached 1, indicating that imbalance persisted at the mRNA level despite the accumulation of βmin-globin mRNA induced by Epo.

Globin chain imbalance is a characteristic feature of β thalassemia. We examined whether the accumulation of βmin-globin mRNAs in treated mice would translate into a more efficient synthesis of βmin-globin chains.

Globin chain synthesis was investigated by metabolic labeling of blood cells (Table 1 and Figure 1F). The βmin/α-globin chain synthesis ratio was 0.69 ± 0.036 before treatment. Values ranged between 0.74 and 0.99 in mice 1 through 7 at 4 weeks after DNA injection and were still higher than before treatment in mice 2 through 4 at 11 weeks. Wilcoxon tests showed that βmin/α-globin chain synthesis ratios were related to the βmin/α-globin mRNA ratio (P = .0117 at weeks 4 and 11), to serum Epo concentrations (P = .0077 at week 4, P = .0117 at week 11), and to Hct (P = .0077 at week 4, P = .0117 at week 11). As for mRNA levels, the ratio never reached 1, indicating that unpaired α chains were still produced. These results showed that the accumulation of βmin-globin mRNA in response to Epo stimulation subsequently stimulates the synthesis of βmin-globin chains.

Unpaired α-globin chains affect erythrocyte quality by liberating iron, binding to membranes, and altering lipids and proteins through oxidative mechanisms. These phenomena result in premature cell destruction. We examined whether increased βmin-globin synthesis subsequent to Epo stimulation allowed for a correction of these defects.

The α-globin chain content was measured in erythrocyte ghosts. Results showed a dramatic reduction at 4 weeks in mice 2 through 7 and a complete disappearance in mouse 1 (Table 1). Changes were minimal in mice 8 and 9. Improvement persisted at week 11 for mice 1 through 3, whereas phenotype reversed to accumulation in other mice. Wilcoxon tests showed a significant relationship between the reduction of α-globin chain content in erythrocyte ghosts and serum Epo concentrations (P = .0077 at 4 weeks,P = .0173 at 11 weeks), βmin/α-globin mRNA ratio (P = .0117 at 4 and 11 weeks) and βmin/α-globin synthesis ratio (P = .0109 at 4 weeks,P = .0117 at 11 weeks).

Spectrin and band 3 are 2 functionally and quantitatively important proteins of erythrocyte membranes. In normal erythrocytes, spectrin α plus β chains and band 3-globin area represent 30.6% ± 5.9% and 22.1% ± 3.5% of total membrane proteins, respectively. Spectrin was decreased in β thalassemic erythrocytes before treatment, whereas the band 3-globin area was slightly increased, so that the ratio of spectrin to band 3 was 0.63 ± 0.06 in β thalassemic mice compared to 1.39 ± 0.14 in heterozygotes. After treatment, the proportions of spectrin α plus β chains increased in mice 2 through 7 and the amount of proteins found in the band 3-globin area decreased (Table 1). This was associated with the disappearance of many protein fractions in the low-molecular-weight range, indicating a general improvement of membrane protein profiles (not shown). Improvement was related to the disappearance of α-globin chains in membranes (P = .0109 at 4 weeks, P = .0117 at 11 weeks) and to the increased amount of βmin-globin mRNA (P = .0117 at 4 and 11 weeks).

A relationship has been described between the presence of surface immunoglobulins and the abnormal arrangement of band 3 proteins in β thalassemic cells.41,42 The proportion of erythrocytes carrying surface immunoglobulin detectable by bead-rosette anti-immunoglubulin assay was determined before and after treatment. Results showed a significant reduction of membrane-bound immunoglobulins in treated mice (Table 1). Reduction was related to that of proteins found in the band 3-globin area (P = .0077 at 4 weeks, P = .0117 at 11 weeks), to the spectrin/band 3-globin ratio (P = .0077 at 4 weeks, P = .0117 at 11 weeks), to the α-globin chain content in membranes (P = .0109 at 4 weeks,P = .0117 at 11 weeks) and to βmin-globin mRNAs levels (P = .0117 at 4 and 11 weeks).

Altogether, these data indicated that the accumulation of βmin-globin mRNA and the increased synthesis of βmin-globin chains in response to Epo stimulation led to the production of qualitatively improved erythrocytes.

Iron decompartmentalization promotes the targeting of auto-oxidative damages to the cell membrane and thereby contributes to β thalassemia pathophysiology.43 

We measured the size of various erythrocyte iron compartments. Pretreatment values illustrated iron overload in β thalassemic cells as compared to heterozygotes. Data were consistent with previous observations made in mouse37 and human44 β thalassemic cells. Values were not significantly different after treatment. In ghost erythrocyte membranes, the accumulation of free iron represented 5.26 ± 0.85 nmoL/mg protein before treatment and 4.92 ± 1.22 at 4 weeks (Figure 2A). Nonheme iron, including both free iron and other nonheme iron deposits such as ferritin iron, was 12.2 ± 4.27 nmol/mg before treatment and 12.13 ± 4.43 at 4 weeks (Figure 2B). Heme iron, including hemoglobin, hemichrome, and free heme, was measured in erythrocyte ghosts and in IOMs. Values found for heterozygotes, untreated, and treated mice (4 weeks) were 1.73 ± 0.09, 2.75 ± 0.72, 2.29 ± 0.63 nmol/mg, respectively, in ghosts, and 0.75 ± 0.06, 1.19 ± 0.36, 0.99 ± 0.35 nmol/mg, respectively, in IOMs (Figure2C).

This study was conducted of exploring the mechanisms governing the effects of Epo overproduction in β thalassemia. Transient Epo secretion was induced in β thalassemic mice by naked DNA electrotransfer into muscles. Globin mRNA content in reticulocytes, globin chain synthesis, and erythrocyte membrane defects were analyzed. As we previously observed, high serum Epo concentrations were strongly related to the appearance of an effective erythropoiesis that improved the β thalassemic phenotype.9 

Real-time quantitative PCR allowed a precise measurement of globin mRNAs in circulating reticulocytes of living animals. The amount of α-globin mRNA was independent of serum Epo concentration, indicating that Epo affects neither the synthesis nor the stability of these transcripts. In contrast, a significant relationship existed between serum Epo concentrations and βmin-globin mRNA content in reticulocytes. This suggests that a consequence of high serum Epo concentrations would be to stimulate either the accumulation of βmin-globin transcripts in erythroid progenitors or the proliferation of erythroid progenitors committed to high βmin-globin gene expression.

The ratio of βmin/α-globin mRNA was directly related to that of globin chain synthesis, both being proportional to serum Epo concentrations. However, βmin-globin mRNA amounts increased more rapidly with serum Epo concentrations than βmin-globin chain synthesis. Thus, translation of βmin-globin mRNA was not facilitated by high serum Epo concentration and the effect of Epo on globin chain synthesis can be accounted for solely by the accumulation of βmin-globin mRNAs. It is therefore unlikely that Epo modulates the translational efficiency of globin mRNAs.

The amount of membrane-bound α-globin chains was inversely proportional to serum Epo concentrations. The α-globin-mRNA amounts, α-globin chain synthesis, and α-globin content (not shown) in erythrocytes did not vary with serum Epo levels. In contrast, membrane-bound α-globin chain levels were inversely related to βmin-globin mRNA copy numbers and to βmin/α-globin chain synthesis. Thus, the amount of unpaired α chains seemed to be solely determined by the available amount of βmin-globin chains. Accelerated synthesis alone can account for increased amounts of βmin-globin chains. It may also be combined with reduced proteolysis, although there is no direct evidence for such an effect of Epo. The observation that the spectrin/band 3 ratio and membrane-bound immunoglobulins were significantly linked to α-chain contents in erythrocyte membranes and thus to βmin-globin chain synthesis levels indicates that Epo effects on membrane protein components were mediated by its action on the accumulation of βmin-globin mRNA.

Taken together, these results indicate that the mechanisms by which Epo overproduction induced an effective erythropoiesis in β thalassemic mice was primarily by favoring the accumulation of βmin-globin mRNA in immature erythroid cells. Consistently, accumulation of β^C globin mRNA in response to intense Epo stimulation was previously described in anemic sheep5 and in sheep bone marrow cultures.17 Drugs that reactivate fetal erythropoiesis in human erythroid cell cultures, such as hydroxyurea, 5-azacytydine and butyrate, increase γ-globin mRNA amounts.16 In contrast, it has been shown that the compensatory increase of βmin-globin chain synthesis in untreated β thalassemic mice affects mRNA translation rather than accumulation.18 We assume that this mechanism was hidden by the accumulation of βmin-globin mRNA in β thalassemic mice secreting high amounts of Epo.

Several lines of evidence suggest that the accumulation of βmin-globin mRNA in C57Bl/6^Hbbth mouse reticulocytes is subsequent to the expansion of an erythroid progenitor compartment in which βmin-globin transcripts are abundant. A similar mechanism could possibly be recruited for a reactivation of fetal erythropoiesis in humans, although effectiveness may require higher Epo serum concentration. In vitro cultures of mouse45 and human10 erythroid progenitors have shown that Epo stimulates the expansion of a cell compartment with a potential for βmin- or γ-globin synthesis. Consistent results were obtained during a short stimulation with rhuEpo in normal or anemic baboons.7 A model has been proposed based on the observation that (1) the distribution of HbF is heterocellular in normal human adult bone marrow with a small percentage of maturing progenitors (F cells) containing HbF46,47 and (2) the most primitive burst-forming unit-erythrocyte, when triggered prematurely to making erythroblasts, propagates an increased proportion of F cells among their progeny.48 The model states that increasing HbF synthesis during stress, thus possibly in response to intense Epo stimulation, involves the accelerated maturation of progenitors, leading to reprogramming or to recruiting erythroid progenitors that will give rise to F cells.49Investigating whether this model may account for anemia correction in β thalassemic mice requires that erythropoiesis is maintained at a steady state, preferentially normocythemia. Indeed, exploration during acute Epo stimulation would not allow us to distinguish the effects observed in β thalassemic mice from those previously described in normal animals.

We obtained robust Epo secretion in β thalassemic mice by the intramuscular injection of naked DNA followed by an electric shock, as previously reported in normal mice.26 Despite a slow decline of Epo serum concentrations over time, this method had a sustained effect on erythropoiesis in normal mice and in rats, resulting in a persistent polycythemia.50 Although serum Epo concentration decreased with a similar kinetics in β thalassemic mice, the effects on erythropoiesis were much more limited in time. Serum Epo concentrations above which effects on erythropoiesis were visible were higher in β thalassemic mice than in normal animals. Whereas threshold values varied depending on β thalassemic animals, they appeared higher than 150 mU/mL. In contrast, normal animals with 32 mU/mL were already polycythemic.26 A likely explanation is that different mechanisms are involved. Whereas induction of polycythemia in normal animals is the consequence of the stimulation of cells committed to adult erythropoiesis, improvement of erythropoiesis in β thalassemic mice supposes the recruitment of cells committed to the accumulation of βmin-RNA. Because effects on erythropoiesis were not maintained over time, we could not attempt to obtain steady-state normocythemia by modulating Epo secretion levels through the tetracycline-dependent expression cassette present in the vector. This would be better performed by using gene transfer vector derived from type 2 adeno-associated virus, which allows sustained Epo delivery levels.9 

It is noticeable that disease correction remained partial even at the peak Epo secretion. Microcytosis and hypochromia persisted, except in animals with intense polycythemia. Iron overload was not corrected in erythrocytes. Persistent dyserythropoiesis, as illustrated by high reticulocyte counts, probably accounted for incomplete phenotypic reversion. We presume that therapeutic strategies for β thalassemia based on Epo delivery will have to find a compromise between the induction of effective erythropoiesis providing some clinical benefit and the persistence of an ineffective erythropoiesis, the complete reversion of which would require the induction of polycythemia.

Despite this intrinsic limitation, the strategy could be of interest for improving erythropoiesis in β thalassemic patients. Electrotransfer of naked DNA is a safe, noninvasive, and low-cost method that could be proposed, possibly in combination with other treatments, to large populations. Although tolerance to high Epo dosages has been documented, transient secretion is certainly advantageous in terms of security, whereas readministration would ensure a sustained effect, if desired. Moreover, control of dose delivery would probably be feasible by modulating doxycycline dosage, if required.

We especially acknowledge Dr W. Hillen and Dr H. Bujard for providing us with the transactivator rtTA2^s-S2 and for their authorization to use it prior to publication. We are grateful to Dr Y. Beuzard and to Dr G. Ciliberto for helpful discussions during the origin of the work. We also thank Dr O. Schwartz for useful comments on the manuscript.