A novel murine model of Cooley anemia and its rescue by lentiviral-mediated human β-globin gene transfer

The β-thalassemias are caused by more than 200 mutations that lead to decreased or absent production of the β chain of hemoglobin (Hb).1-3 Most of the mutations are point mutations in the β-globin gene that either decrease the level of transcription or mRNA stability or lead to nonfunctional mRNA, as is the case in nonsense and frameshift mutations. Other mutations include deletions in the β-globin gene, as well as rare deletions located distally to the β-globin gene itself.1,3,4 There is a large spectrum in the severity of the disease found in homozygotes and compound heterozygotes. Patients affected by β-thalassemia intermedia do not require chronic transfusion therapy in the first years of life, but they often worsen over time, eventually developing hypersplenism, osteopenia, extramedullary hematopoiesis (EMH), and iron overload. Some patients eventually require splenectomy and blood transfusions. In patients with β-thalassemia major, or Cooley anemia, the absent or extremely reduced production of the β chain of Hb causes severe ineffective erythropoiesis, massive erythroid hyperplasia in the bone marrow and extramedullary sites, and hemolysis. The ensuing iron overload can lead to endocrine deficiencies, cirrhosis, and cardiac failure. In the absence of lifelong transfusions, the disease is lethal.5 Current disease management consists of prenatal diagnosis, transfusion therapy, or allogeneic bone marrow transplantation.1,3,4 Only the latter is curative, but this option is limited to a minority of patients for whom a histocompatible donor can be identified.

New approaches aimed at ameliorating the condition of patients with severe β-thalassemia are greatly needed. Genetic approaches based on the transfer of a regulated human β-globin gene in autologous hematopoietic stem cells (HSCs) represent an attractive potential treatment.6 Their implementation has been hampered by the difficulty of appropriately regulating expression of globin genes in the progeny of virally transduced HSCs. The therapeutic globin gene must indeed be expressed in a tissue-specific fashion and at a high level.7,8 We recently demonstrated that complex lentiviral vectors harboring the human β-globin gene and an optimized combination of proximal and distal transcriptional control elements direct erythroid-specific and sustained globin gene expression in murine bone marrow chimeras.9 In mice with β-thalassemia intermedia, the TNS9 lentiviral vector durably ameliorated anemia.9,10 A similar vector encoding a variant globin gene reduced sickle cell formation in 2 murine models of sickle cell disease (SCD).11 Obtained in hitherto available models of severe hemoglobinopathies, these encouraging findings do not address whether stem cell–based gene therapy is potentially curative for a deficiency as profound as that found in Cooley anemia.

Three mouse models of β-thalassemia have been generated to date. These mouse models are affected by different degrees of β-thalassemia intermedia. The th1 model results from the deletion of the β^major gene; Hbb^th1/th1homozygotes exhibit a moderate form of thalassemia.12 A second model (th2) was generated by insertional disruption of the β^major gene, causing lethality in Hbb^th2/th2 homozygotes but only a very mild phenotype in the heterozygotes.13 The third model,th3, was generated by deletion of both the β^major and β^minor genes.14Mice homozygous for this deletion die late in gestation, while heterozygotes are viable and thalassemic. Adult Hbb^th1/th1 and Hbb^th3/+mice exhibit anemia (slightly more severe in the latter model), abnormal red cell morphology, and splenomegaly, and develop spontaneous hepatic iron deposition similar to that found in humans with β-thalassemia intermedia. The lack of an animal model for Cooley anemia has limited the full investigation of the pathophysiology underlying this disease and hampered the evaluation of both pharmacologic and genetic treatments.

Here we describe a severe anemia model in which mice engrafted with β-globin–null (Hbb^th3/th3) fetal liver cells succumb to ineffective erythropoiesis within 60 days. This novel model enabled us to investigate whether globin gene transfer could rescue a massive protein deficit. We show that treatment with the TNS9 vector durably rescues most treated chimeras. In hematopoietic chimeras with a mean vector copy number of 1.0 to 2.4 in hematopoietic and blood cells, stable production of chimeric Hb (Hbb^hu, consisting of chimeric murine β₂:human β₂ tetramers) was achieved, reaching levels up to 12 g/dL and effectively treating the anemia. These findings strongly support the feasibility of a genetic treatment for β-thalassemia and have general implications for devising stem cell–based treatments for severe blood disorders.

Donor fetal liver cells (FLCs) were harvested from embryonic day (ED) 14.5 embryos obtained by intercrossing Hbb^th3/+ mice (Jackson Laboratories, Bar Harbor, ME). Embryo genotypes were screened by analyzing Hb by cellulose acetate electrophoresis (Figure 1A). FLCs were resuspended in X-VIVO-15 serum-free medium and supplemented with 10 ng/mL IL-1α, 100 U/mL IL-3, 150 U/mL IL-6, 10 ng/mL Kit ligand obtained from Genzyme (Cambridge, MA), 0.5 mM β-mercaptoethanol obtained from Sigma (St Louis, MO), 200 mM L-glutamine, 100 IU/mL penicillin, and 100 μg/mL streptomycin. FLCs were then pelleted and resuspended in serum-free medium containing concentrated lentiviral supernatant and supplemented with 4 mg/mL polybrene obtained from Sigma, 200 mM L-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin, and cytokines as above, and cultured for 8 hours. Transduced FLCs (2 × 10⁶) were then intravenously injected into each of the irradiated female recipients to establish bone marrow chimeras. Recipient mice (11- to 14-week-old C57/BL6 mice) were irradiated with 12 Gy (split dose 2 × 6 Gy) on the day of transplantation. Vector stocks were generated by triple transfection into 293T cells, as previously described.9 Concentrated viral stocks with titers of 1-2 × 10⁸ infectious particles/mL were used to infect 2 × 10⁶ FLCs in a final volume of 1 mL (multiplicity of infection = 50). The lentiviral vectors TNS9 and pHR'eGFP (enhanced green fluorescent protein), derived from pHR′LacZ,15 have been previously described.9 

Blood samples were obtained by retro-orbital puncture under ether anesthesia. Total hemoglobin levels, red cell counts, hematocrit levels, neutrophil counts, platelet counts, and reticulocyte counts were measured on a blood cell count analyzer (H System, Bayer, Leverkusen, Germany). Plasma erythropoietin (EPO) concentrations were measured using an enzyme-linked immunosorbent assay (ELISA) kit for human EPO (R&D Systems, Oxon, United Kingdom).

Red cell lysates from freshly collected peripheral blood were analyzed by cellulose acetate electrophoresis obtained from Helena Laboratories (Beaumont, TX). Hemoglobin bands were visualized by Ponceau S staining and quantitated by densitometry, as previously described.9 

For quantification of vector copy number in HSC chimeras, genomic DNA from tissues was isolated, digested with BamHI, and studied by Southern blot analysis using a [³²P]dCTP-labeled NcoI/BamHI fragment of the human β-globin gene as probe. TNS9-transduced single-copy murine erythroleukemia (MEL) clone controls also were digested with BamHI. Radioactive bands were quantitated by phosphorimager analysis (Molecular Dynamics).

Total RNA was extracted from bone marrow (BM), spleen, and blood using TRIzol (Gibco-BRL). Quantitative primer extension assays were performed as previously described,9 using the Primer Extension System-AMV Reverse Transcriptase kit (Promega) with [³²P]dATP end-labeled primers specific for lentiviral-encoded human β-globin (5′-CAGTAACGGCAGACTTCTCCTC-3′), mouse β-globin (5′-TGATGTCTGTTTCTGGGGTTGTG-3′), and mouse β-globin (5′-CCTTGATGTTGCTTTTGTCTTC-3′) with predicted extension products of 90 bp, 53 bp, and 69 bp, respectively. Radioactive bands were quantitated by phosphorimager analysis (Molecular Dynamics). All measurements were standardized to RNA samples obtained from YAC transgenic mice harboring one copy of the human β-globin gene (line A85.68,16) as previously described.9 

Four-micron tissue sections were stained with hematoxylin and eosin or with a Gomori iron stain and examined under light microscopy, as previously described.10 Slides of control and treated mice were assessed in a blind manner.

For statistical analyses, we used the permutation rank sum statistic to determine whether hematologic parameters differed between treated and mock-treated groups. A low P value is evidence that the 2 populations are different. The Wilcoxon rank sum statistic was used to compare survival curves between groups.

Homozygote mice lacking adult β-globin genes (Hbb^th3/th3) die late in gestation.14 We therefore attempted to adoptively transfer the hematologic deficit to adult mice by transplanting FLCs harvested from screened ED 14.5 embryos (Figure 1A). Congenic C57BL/6 recipient mice were irradiated (12 Gy) before intravenous infusion of 2 × 10⁶ FLCs. Mice engrafted with Hbb^+/+ or Hbb^th3/+ FLCs survived for at least 8 months, corresponding to the duration of this study (n = 11 and 24, respectively; Figure 1B). In contrast, recipients of Hbb^th3/th3 cells showed a progressive decrease of body mass and activity, eventually requiring termination 7 to 9 weeks after transplantation (T₅₀ = 50 days, n = 31). These mice died significantly later than radiation controls (T₅₀ = 15 days, n = 10,P < .01), suggesting that the cause of death was neither failure to engraft nor pancytopenia.

Hematologic analyses performed 6 to 8 weeks after transplantation in mice engrafted with Hbb^th3/th3 FLCs revealed severe anemia (2.8 ± 0.8 g/dL of Hb levels, versus 13.2 ± 1.0 in 11 Hbb^+/+ chimeras and 11.1 ± 2.1 in 24 Hbb^th3/+ chimeras). Low red blood cell (RBC) counts, hematocrit values, and reticulocyte counts, together with high levels of serum erythropoietin (S.R., M.S., unpublished observations), further confirmed the development of a profound erythroid deficiency. Platelet counts (Figure2), as well as neutrophil counts (data not shown), were comparable in experimental and control groups, excluding aplasia or graft failure as the cause of death. Analyses of murine globin transcripts in BM, spleen, and blood by primer extension analysis showed that murine β-globin (muβ) expression was extremely low or undetectable in bone marrow, spleen, and blood of mice engrafted with Hbb^th3/th3 FLCs (data not shown). The presence of the β-globin transcript further confirmed that these chimeras developed a selective erythroid defect caused by the lack of β-globin expression. The profound anemia settled in after 50 days, consistent with the clearance rate of the recipient's normal RBCs.17 

After 6 to 7 weeks, mice engrafted with Hbb^th3/th3 FLCs exhibited massive splenomegaly, far exceeding that found in mice engrafted with Hbb^th3/+ FLCs (Figure 1C). The massive splenomegaly in the Hbb^th3/th3 chimeras was due to marked erythroid hyperplasia consisting mostly of immature nucleated erythroid cells that obliterated the normal microscopic architecture of the spleen. Mature RBCs, still present in the Hbb^th3/+ chimera albeit in reduced numbers compared with Hbb^+/+ chimera, were virtually absent in their bone marrow and spleen (Figure 3A-B). EMH was absent in the liver of Hbb^+/+ chimeras, moderate in Hbb^th3/+ chimeras, and extensive in Hbb^th3/th3 chimeras (Figure 3C). We also examined hepatic iron accumulation, which is a major characteristic of the systemic disease found in patients with severe thalassemia, arising as a consequence of hemolysis, ineffective erythropoiesis, and increased gastrointestinal iron uptake. In Hbb^+/+ and Hbb^th3/+ chimeras, iron accumulation was, respectively, absent and moderate, predominantly within Kupffer cells (Figure 3D). In the Hbb^th3/th3 chimeras, iron deposition in the liver was found as early as 6 weeks after transplantation, with preferential accumulation in the cytoplasm of the hepatocytes around the portal tracts (Figure 3D). Altogether, these pathologic findings confirmed the presence of severe ineffective erythropoiesis, extensive EMH, and rapid iron accumulation.

The establishment of a novel adult model of severe β-thalassemia allowed us to investigate the efficacy of lentivirus-mediated globin gene transfer in correcting the lethal phenotype and the anemia. Hbb^th3/th3 as well as control Hbb^+/th3 and Hbb^+/+ FLCs were transduced with either the TNS9 vector or a control lentiviral vector encoding eGFP. All mice engrafted with eGFP-transduced Hbb^th3/th3 FLCs (n = 20) died as expected within 60 days (Figure 4). In contrast, all mice engrafted with TNS9-transduced Hbb^th3/th3 FLCs (n = 13) survived at least 4 months (P < .01, TNS9 versus eGFP-transduced Hbb^th3/th3; Figure 4), thus confirming the deficit in β-globin synthesis to be the cause of death. These mice showed a remarkable improvement of their hematologic parameters (Figure 2) 8 weeks after transplantation, with Hbb^hu accounting for 30%-65% of total Hb. Of the 13 mice TNS9-treated Hbb^th3/th3 chimeras, 10 stably produced Hb consisting predominantly or exclusively of Hbb^hu. Over 8 months (Figure 5A-B), 6 consistently showed levels of Hbb^hu above 95% of total Hb levels, therefore surviving on the sole basis of sustained production of lentivirus-encoded human β-globin. Chimeras with more than 5% murine Hb were excluded from further analyses to rule out any contribution of residual host erythropoiesis to their pathology and survival. Secondary bone marrow chimeras derived 5 and 6 months after transplantation from 4 of the 6 healthy mice with more than 95% Hbb^hu expressed high levels of the human β-globin levels in peripheral blood after 2-4 months (data not shown). These findings, consistent with our results obtained in normal mice,9 indicated that TNS9-treated chimeras were rescued by transduced fetal liver HSCs.

Our long-term studies focused on chimeras with less than 5% residual murine Hb levels in the 8 months following transplantation (n = 6). Over this period the mice showed stable and long-term expression of Hbb^hu, averaging 6.5 ± 2.9 g/dL. This level approximates that of hemizygous Hb production (8.1 ± 0.3 g/dL in Hbb^th3/+ chimeras, Table1). Pathologic examination performed between 5 and 8 months after transplantation showed variable degrees of ineffective erythropoiesis, commensurate to the degree of anemia. Tissue sections from 2 mice, one with a level of 4.7 g/dL Hb and the other with a level of 12 g/dL are shown in Figure6. The findings in the latter are similar to the eGFP-transduced Hbb^+/+ control (Figure 3), except for small and rare foci of EMH in the liver (Figure 6 right column). Virtually no iron deposition was detected. The long-term chimera with 4.7 g/dL Hb levels showed features intermediate between control Hbb^th3/th3 and Hbb^th3/+ chimeras (Figure 6 left column). To further demonstrate the restoration of erythroid maturation, we analyzed splenic erythroid populations by fluorescence-activated cell-sorter scanner (FACS) analysis. In control eGFP-Hbb^+/+ chimeras, the majority (72% ± 7%) of Ter-119⁺ cells corresponded to erythrocytes and reticulocytes,18 shown in the top left box of each quadrant in Figure 7. This fraction decreased to 52% ± 13% in eGFP-transduced Hbb^th3/+ chimeras and was virtually absent in recipients of eGFP-Hbb^th3/th3 FLCs. In TNS9-treated Hbb^th3/th3 chimeras, this fraction was restored to 25% ± 8% (n = 4), thus directly demonstrating restoration of erythroblastic maturation.

The thorough evaluation of globin gene transfer as a potential treatment for severe hemoglobinopathies requires murine models of relevant severity in which the impact of gene transfer efficiency, transgene expression, chimerism, and the recipient's genetic background can be evaluated. Using the TNS9 vector, we previously demonstrated efficient gene transfer in bone marrow hematopoietic stem cells, tissue-specific transgene expression, and long-term correction in a mouse model of β-thalassemia intermedia.9,10 In order to fully evaluate the therapeutic potential of TNS9 and its potential clinical applicability to the most severe hemoglobinopathies, it is essential to investigate its efficacy in the context of a fatal anemia. Here we address this requirement by establishing a novel mouse model for the most severe form of β-thalassemia, Cooley anemia. We demonstrate that mice die as a consequence of extreme ineffective erythropoiesis and that they can be rescued and cured by lentivirus-mediated transfer of the human β-globin gene.

In this new model, a severe anemia gradually develops, reaching 2-3 g/dL of Hb levels within 6 to 9 weeks. These kinetics are consistent with the disappearance of host RBCs17 and the inability to generate functional RBCs from the graft. Moreover, these mice rapidly develop massive EMH and iron overload, features that are characteristic of severe β-thalassemia. This model should therefore prove useful to investigate novel treatments for congenital anemias, including pharmacologic and genetic approaches. Genetic approaches based on vectors encoding β-like globin genes, either β, γ, or mutants thereof, can be thoroughly tested in this extreme form of β-thalassemia. Lacking expression of any adult endogenous β-globin, this model provides a gold standard for quantifying globin transgene expression and comparing therapeutic vectors. By introducing selected genomic regions of the human β-like globin gene domain, in particular the γ-globin region, pharmacologic treatments for reactivating human fetal hemoglobin19-21 also could be investigated in this adult model of disease.

Our results establish lentivirus-mediated gene transfer as an effective, expeditious, and broadly applicable approach to study and rescue lethal hematopoietic phenotypes in animal models. In the absence of any selection, fetal liver stem cells were efficiently transduced with a lentiviral vector encoding a regulated, tissue-specific gene. Complex vectors like TNS9 provide a powerful tool for rescuing lethal phenotypes obtained by targeted gene disruption as well as a novel approach for investigating gene regulation. In contrast to the multicopy concatamers typically found in transgenic mice following pronuclear DNA injection,22 lentivirus-mediated gene transfer will be of great use in gene expression studies by enabling titration of gene delivery and thus facilitate the generation of single-copy mice. The combination of highly efficient gene transfer,23 the ability to package large genomic elements and thus enhance gene regulation,9 and the preferential integration into sites of active chromatin24 make lentiviral vectors a major tool for experimental gene transfer. The same vectors can be used to either generate hematopoietic chimeras or transgenic animals through infection of embryonic stem cells.25,26 As shown here, fetal liver cell transduction is highly efficient, enabling rapid and effective hematopoietic reconstitution and thus providing a significant practical advantage over transgenesis.

Using this model, we could determine for the first time the absolute level of human β chain production obtained in vivo in the absence of competing murine β chains. Unlike transgenic mice, hematopoietic chimeras harbor variable mixtures of transduced and nontransduced cells, with the former also varying in their vector copy number. Chimeras with endogenous repopulation in excess of 5% were not included in these analyses to rule out any contribution of residual host erythropoiesis to their pathology and survival. In the 6 long-term chimeras with negligible residual host erythropoiesis (< 5% Hbb^mu), the mean Hb levels were 6.5 ± 2.9 g/dL, in conjunction with an average vector copy number of 1.6 ± 0.6. In one chimera showing stable production of 12 g/dL Hbb^hu, the vector copy number was 2.2-2.4 in the different hematopoietic tissues. This mouse was virtually cured of anemia. The persistent, moderate EMH noted in this mouse, greater than expected for this Hb level,10 suggests that the chimeric Hbb^hu may not be functionally equivalent to Hbb^mu.

The TNS9 vector encodes the human β-globin gene, including an extended promoter and 2 proximal enhancers, 3 large genomic fragments derived from the human β-globin locus control region, and additional synthetic GATA-1 transcription factor binding sites. Globin transcriptional control elements have been extensively studied,27,28 without conclusively defining what minimal segments and what combination thereof are best suited to generate a miniature β-globin transcription unit. The relative roles of the locus control region and of other genetic elements spanning the human β-globin locus and its chromatin domain in regulating globin gene expression remain controversial.29-31 The combination adopted within the TNS9 vector results from a systematic analysis of numerous combinations (TNS 1 through 9; S.R., C.M., M.S., unpublished observations). As shown here, most Hbb^th3/th3 chimeras were rescued by TNS9. This unequivocally demonstrates that a major therapeutic benefit can be achieved in a lethal hemoglobinopathy. Using a vector closely reproducing the structure of TNS9, termed β89, which differs in one amino acid in the β-globin gene, the promoter region (−615 from the cap site in TNS9 versus −266 in β89) and boundaries of the locus control region (LCR) elements (3.2 kb in TNS9 versus 2.7 kb in β89), Pawliuk et al achieved inhibition of red blood cell dehydration and sickling with correction of hematologic parameters and splenomegaly11 in 2 murine SCD models, Berkeley32 and SAD.33 This result was obtained with an average of 3 copies of the vector per cell.11 It is unclear whether such a high vector copy number is required to ameliorate anemia or increase lifespan in these models of SCD and whether this high vector copy number is also necessary to treat murine β-thalassemia intermedia or β-thalassemia major. It would be interesting to investigate β89 and other globin vectors in the present model of β⁰-thalassemia to determine what mean vector copy number is sufficient to rescue lethal anemia. Although seemingly minimal, the differences between TNS9 and β89 may well result in different levels of β-globin production and different susceptibility to position effects. In consideration of the potential risk of insertional oncogenesis associated with randomly integrated transgenes, it is essential in our view to optimize transgene expression so as to achieve therapeutic benefits with the lowest possible vector copy number.

Based on the results obtained in the model of Cooley anemia described here, we conclude that the TNS9 lentiviral vector expresses high enough levels of human β-globin to warrant its clinical evaluation. In addition to requiring elevated and regulated globin gene expression, the prospective clinical use of lentiviral vectors requires that other efficacy and safety features be ascertained. At present, the efficacy of lentiviral vectors in transducing human HSCs appears to be at least as good as that of oncoretroviral vectors. This matter is still under investigation,34-42 including, importantly, marking studies in nonhuman primates.43-45 The transduction efficiency of lentiviral vectors would have to prove to be vastly superior to that of oncoretroviral vectors to justify their clinical investigation for this reason alone. In the case of the severe hemoglobinopathies, the unparalleled efficacy of lentiviral vectors in permitting stable transfer of large genomic fragments and regulation of β-globin expression provides a powerful argument in favor of their use.46 The major safety concerns relate to the risk of generating a replication-competent lentivirus (RCL) and to that of insertional oncogenesis. The former depends on the establishment of safe packaging cell lines and dependable RCL detection assays. The hitherto positive experience with oncoretroviral vectors47provides an encouraging precedent for the clinical investigation of lentiviral vector systems48,49 and suggests that the generation of safe, stable, and high-titer lentiviral packaging cell lines should be feasible.

While insertional mutagenesis is unavoidable with retroviral vectors—as for any gene transfer method resulting in random integration—the risk of insertional oncogenesis is difficult to apprehend. Although insertional oncogenesis is deemed to be a rare event,50 2 recent reports51,52 raise the concern that its frequency may have been underestimated. In both instances, genes encoding molecules involved in the regulation of cell survival, ΔLNGFR51,53 and γ_c,52,54 were transduced in bone marrow cells under the nonspecific transcriptional control of the retroviral long terminal repeat (LTR). In both cases, integration was adjacent to oncogenes, respectively Evi-1 and LMO-2. The complete analysis of the transforming events associated with leukemogenesis in both instances is still pending and therefore preclude definitive conclusions. Importantly, lineage-restricted vectors with U3-deleted LTRs provide in principle a mean to reduce the risk of insertional oncogenesis.55 Globin vectors encompassing the β-globin promoter are transcribed only in late stages of proerythroblastic differentiation, prior to nuclear extrusion, except for possible low-level “leakiness” at some integration sites. While the risk of gene disruption is unchanged relative to LTR-driven vectors, the risk of oncogene activation would therefore be confined to a single lineage. The risk of promoter transactivation by vector-encoded enhancers could be further reduced by the incorporation into the vector of insulators or other elements with enhancer-blocking activity.56,57 A comprehensive definition of the risk of insertional oncogenesis will require extensive studies over time. Ultimately, like any other medical decision, therapy will be best guided by a rational risk-benefit assessment. Based on our prior reports9,10 and the present one, we conclude that lentiviral vectors are very promising drugs for the treatment of β-thalassemia intermedia and β-thalassemia major.

We thank Ping Zhu, Ellinor Peerschke, and Diane Magravati for technical assistance and Katerina Politi for reviewing the manuscript.