Modulating erythrocyte chimerism in a mouse model of pyruvate kinase deficiency

Although many inherited blood disorders are curable through stem cell transplantation, the applicability of the procedure is limited in part by its toxicity. One way to reduce the toxicity associated with stem cell transplantation is through the use of attenuated conditioning regimens that fall short of producing complete myeloablation and lead to the establishment of donor/host chimerism. Studies in inherited red blood cell (RBC) disorders have shown that relatively small percentages of normal donor stem cells can substantially ameliorate the severity of these diseases. Patients in whom chimerism has developed after stem cell transplantation for β-thalassemia or sickle cell disease demonstrate that as few as 10% of normal donor progenitors can functionally compensate for the anemia that characterizes these disorders.^1-3  Similar conclusions pertain to the relevant mouse models of these disorders.^4-9 

However, chimerism alone does not correct all the manifestations of hemolysis. Transplantation studies in a sickle cell mouse model demonstrated that hematologic and pathologic correction required 100% peripheral RBC chimerism.¹⁰  In studies of hematopoietic chimerism in a dog model of pyruvate kinase (PK) deficiency, improvement in the anemia failed to eliminate persistent hemolysis in 1 animal and resulted in severe iron overload and cirrhosis.¹¹  After stem cell transplantation for thalassemia, persistent iron overload is treated with aggressive phlebotomy to forestall progression of liver fibrosis.¹²  However, patients with chimerism may not be able to tolerate phlebotomy because of low hematocrits, necessitating treatment with chelation therapy.^3,13  Methods that would allow for the erythropoietin-independent expansion of donor-origin erythropoiesis may suppress erythropoietin production and, in turn, residual aberrant erythropoiesis. Additionally, expanding donor hematopoiesis in vivo may aid in preventing the graft rejection that plagues nonmyeloablative transplantation for hemoglobinopathies.¹⁴ 

We have previously described a system that uses small-molecule drugs as artificial growth factors for genetically modified cells. These drugs, called chemical inducers of dimerization (CIDs), exert their activity by directing the self-association of hybrid signaling molecules composed of a CID-binding domain and the intracellular portion of a growth factor receptor.^15,16  We have shown that CID-mediated activation of the intracellular portion of the thrombopoietin receptor, mpl, can induce growth of genetically modified hemopoietic cells from mice, dogs, and humans.^17-19  On in vivo administration, CID-induced activation of mpl induces a dramatic expansion of genetically modified RBCs. In this study, we have used a mouse model of pyruvate kinase (PK) deficiency to test whether CIDs can be used to augment the contribution of normal donor erythrocytes in patients with mixed chimerism. The PK-deficient mouse (CBA-Pk-1^slc/Pk-1^slc) closely reproduces the disease found in the canine model and in humans,^20-22  exhibiting a phenotype that includes anemia, reticulocytosis, and splenomegaly. Marrow cells from the congenic founder strain, CBA/N, can be used in transplantation experiments to completely correct the disorder.²¹  We show that the selective, CID-mediated expansion of normal donor RBCs can improve inherited anemia and reduce abnormal erythropoiesis.

The retroviral vector F36Vmpl^GFP has been described in a previous publication.¹⁸  Briefly, this vector is based on murine stem cell virus (MSCV)²³  and expresses a single transcript that encodes the CID-responsive cytoplasmic portion of the thrombopoietin receptor (mpl) and the enhanced green fluorescence protein (GFP) (Figure 2A). A point mutation was introduced into the FKBP domain that changed a phenylalanine to a valine at amino acid position 36, allowing for binding of the CIDs AP1903 and AP20187.^19,24  A stable ecotropic producer line (GP+E86) with a titer of 1 × 10⁶ was used for coculture-based gene transfer.

Breeding pairs of the wild-type strain CBA/N and the pyruvate kinase mutant strain CBA-Pk-1^slc/Pk-1^slc were obtained from the SLC facility in Japan.²¹  They were maintained in a specific pathogen-free facility. In addition, the animals were maintained on a breeder diet to limit the development of folate deficiency.

Twelve- to 16-week-old donor mice were treated with 150 mg/kg 5-fluorouracil (5-FU) by intraperitoneal (IP) injection 2 days before bone marrow mononuclear cells were harvested. Cells were prestimulated for 48 hours in Dulbecco minimal essential medium containing 16% fetal calf serum (FCS), 5% interleukin-3 (IL-3)-conditioned medium, 100 ng/mL recombinant human IL-6, and 50 ng/mL recombinant murine stem cell factor (SCF) ina37°C, 5% CO₂ incubator. Cells were then transferred to irradiated (1500 cGy) producer cells and cocultivated using identical growth conditions except for the addition of polybrene (8 μg/mL). Marrow cells were harvested after 48 hours and were immediately infused by tail vein injection. Eight- to 12-week-old recipient mice received variable doses of radiation from a dual cesium Cs 137 gamma source (GammaCell 40; AEC, Kanata, ON, Canada) before infusions ranging from 2 to 10 × 10⁶ transduced bone marrow. Gene transfer was monitored by methylcellulose plating and by scoring progenitor colonies for GFP expression. All mice that underwent transplantation in this study are listed in Table 1, as are mice that underwent similar transplantations and were followed up long term for adverse events.

The drug used to regulate cell expansion was AP20187 (ARIAD Pharmaceuticals, Cambridge, MA). This drug was designed to specifically bind to 2 mutated FK1012 binding proteins (FKBPs). Lyophilized AP20187 was solubilized in 100% ethanol to produce a 5-mg/mL stock that was stored at -20°C. From this stock solution, AP20187 was diluted fresh on the day of injection. The final solution for injection contained 10% polyethylene glycol (PEG) 400 and 1.4% Tween 80. We administered AP20187 by daily intraperitoneal injection in a total volume of 400 μL per injection.

RBC number, reticulocyte percentage, and RBC and platelet GFP expression were determined every 1 to 2 weeks through a 10-μL sample of blood taken from a tail vein. RBC numbers were determined on a Coulter Z series after a 1:50 000 dilution in isotonic saline. Reticulocyte percentage was determined by staining peripheral blood with new methylene blue (Sigma, St Louis, MO). Reticulum-positive cells were counted, and at least 500 cells were scored per sample. Flow cytometry was performed using FACScan (Becton Dickinson, Mountain View, CA). RBCs and platelets were identified based on forward- and side-scatter characteristics. Statistical analysis of groups (with and without CID) used a 2-tailed t test assuming equal variance (Microsoft Excel Statistical analysis package).

Mice were killed and serum was collected. Levels of serum erythropoietin were measured using a murine erythropoietin-specific enzyme-linked immunosorbent assay (ELISA), as described.²⁵  Briefly, monoclonal rat antimurine erythropoietin antibody (BD BioSciences, Mississauga, ON, Canada) was used as the capturing antibody, and polyclonal rabbit antihuman erythropoietin antibody (R&D Systems, Minneapolis, MN) was used as the detecting antibody. Reactivity was revealed using ABTS substrate (Boehringer Mannheim-Roche, Laval, PQ, Canada), and optical density values were read at a wavelength of 405 nm with 492 nm as the reference. Concentrations of erythropoietin in serum samples were calculated against the standard curve generated using recombinant murine erythropoietin (Boehringer Mannheim-Roche).

The following primers were used for polymerase chain reaction (PCR) across the region of the lr-pk gene that contains the single base pair change that causes PK deficiency in this animal model: 5′ primer, ATCTTTTGGATCTGCGCTTC; 3′ primer, GCCAAGAAAACCTTCTCTGCT. The resultant product is 367 bp in length. The mutation results in the loss of a BstEII site. The PCR product from the CBA/N lr-pk gene can be digested with BstEII to result in fragments of 326 and 41 bp. DNA was purified from the peripheral blood of mice that underwent transplantation and was used as the template for the PCR reaction. After digestion with BstEII, the DNA fragments were resolved on 2% agarose. Kodak DS 1D software (Eastman Kodak, Rochester, NY) was used to assign relative fluorescence intensities to the bands and to determine ratios within samples. Values were adjusted to compensate for the differences in band size.

We first determined the relative level of donor chimerism required to improve the anemia associated with pyruvate kinase deficiency. Marrow cells from the CBA/N mouse and the mutant CBA-Pk-1^slc/Pk-1^slc were subjected to the conditions necessary for gene transfer, mixed in various proportions, and infused into lethally irradiated CBA/N recipients. Although 5% of normal donor marrow cells were insufficient to correct the anemia, 10% resulted in near normal RBC numbers (Figure 1).

We next identified a minimal conditioning regimen for the CBA-Pk-1^slc/Pk-1^slc mice that would allow for the engraftment of transduced marrow cells of normal donor origin. Radiation doses as low as 100 cGy have been shown to improve engraftment of donor stem cells.^26,27  However, studies have demonstrated that retroviral transduction impairs stem cell engraftment.^28-30  Therefore, we compared 2 doses of gamma radiation, 100 cGy and 400 cGy, as conditioning regimens for achieving stable engraftment of transduced donor stem cells.

CBA-Pk-1^slc/Pk-1^slc mice were conditioned with either 100 cGy or 400 cGy gamma irradiation in a single dose on the day of transplantation (schematic; Figure 2B). Marrow cells from the CBA/N donors were transduced with the MSCV-based retroviral vector F36Vmpl^GFP (Figure 2A) using a 4-day coculture method that resulted in gene transfer into 49% to 67% of IL-3-responsive colony-forming cells (data not shown).¹⁸  Ten million posttransduction donor cells were injected per recipient. Blood (5-10 μL) was sampled from the tail veins of the mice every 1 to 2 weeks to analyze GFP expression in erythrocytes and platelets, changes in the peripheral blood erythrocyte number, and effects on the percentage of reticulocytes. As shown in Figure 2C, GFP-positive cells disappeared from the blood of mice conditioned with 100 cGy irradiation. By 12 weeks after transplantation, GFP-positive RBCs were no longer evident in the peripheral blood, and GFP-positive platelets were not detected at any time point after transplantation (Figure 2D). These findings suggested that 100 cGy allowed for only brief short-term engraftment resulting in a transient production of GFP-positive donor RBCs. This loss of transduced RBCs might have resulted from a failure to engraft or from an immune-mediated rejection of the transduced GFP-positive cells.³¹ 

In contrast, mice conditioned with 400 cGy exhibited long-term engraftment (more than 40 weeks) with transduced CBA/N stem cells, as evidenced by the continued presence of GFP-positive erythrocytes and platelets. Differences in the percentages of GFP-positive erythrocytes compared with platelets are attributable to the prolonged survival of the CBA/N erythrocytes relative to their PK-deficient counterparts. Mice conditioned with 400 cGy had long-term improvements in RBC numbers (Figure 2E), decreased reticulocytes (Figure 2F), and near-complete normalization in spleen size (Figure 2G.) In contrast, mice conditioned with 100 cGy were indistinguishable from CBA-Pk-1^slc/Pk-1^slc mice.

We next determined whether conditioning with 400 cGy could allow a donor graft to be established with fewer transduced marrow cells. Mice underwent transplantation with a range of transduced donor cells. Figure 3 shows the results using either 2 or 4 million transduced donor cells. Figure 3A shows the percentage of GFP-positive erythrocytes and platelets in 3 mice that underwent transplantation with 4 million donor cells. All 3 mice maintained transduced cells past 20 weeks after transplantation. In addition, the transduced cells were responsive to CID, as evidenced by the increase in GFP-positive RBCs in the treated mouse (Figure 3A, solid squares). Three mice that received 2 million cells did not establish donor grafts, as evidenced by the loss of GFP expression in erythrocytes and platelets (Figure 3B), and CID administration failed to induce the reappearance of GFP-positive cells in these mice.

We expected that anemic mice that had undergone reduced-intensity transplantation with F36Vmpl^GFP-transduced normal marrow cells would respond to CID administration with an increase in RBC numbers. Ten CBA-Pk-1^slc/Pk-1^slc mice irradiated with 400 cGy underwent transplantation with 5 million transduced CBA/N marrow cells. GFP-positive cells from the CBA/N donor persisted long term in all the mice. Initially, all mice showed near normalization of RBC numbers, likely because of the engraftment of short-term repopulating cells from donor CBA/N marrow. By 23 weeks after transplantation, all mice were anemic (RBCs, 7.87 ± 0.69 SEM). At that point the mice were divided into 2 groups, matched for RBC numbers and reticulocyte counts. One group was treated with CIDs, and the control group received injections of the carrier alone without CIDs. The mice that did not receive CIDs had a slow, steady worsening of anemia (Figure 4D), similar to the decline in hematocrit observed in thalassemic mice that underwent transplantation with transduced marrow cells.³²  Initial treatment using 5 days of drug every 4 weeks led to clear increases in GFP-positive RBCs with variable responses in individual mice. In general, mice with low percentages of GFP-expressing blood cells were more anemic and exhibited less pronounced responses to CID treatment. At 36 weeks, the frequency of drug administration was increased to 5 days every other week, resulting in a more uniform increase in GFP-positive RBCs (Figure 4A). Platelets showed a modest response to CID treatment (Figure 4B). The reticulocyte percentage decreased in the animals treated with CID (Figure 4C), except for transient increases in reticulocytes immediately after CID administration. There were significant differences in reticulocyte percentages (P < .01) between CID-treated and untreated mice at 30 and 42 weeks. Mice treated with CID had significantly increased RBC number (Figure 4D, G). The average RBC number after week 30 was 6.39 ± 0.88 × 10⁹/mL for control animals and 9.09 ± 0.67 × 10⁹/mL for animals treated with CID (P < .01). Figure 4G demonstrates that with more frequent administration of CIDs, the treated group developed clear improvements in RBC number compared with untreated mice. The sixth CID dose was withheld for 2 mice because of elevated RBC numbers (more than 13 × 10⁹/mL) (Figure 4, gray arrow). After CIDs were withheld, the percentages of GFP-positive RBCs decreased, with a half-life of 56 days (data not shown). The relatively slow decay rate of transduced donor RBCs was likely caused by the relatively long lifespan of this cell population combined with the persistence of CID-induced RBC production for 1 to 2 weeks after CID withdrawal, as we have noted previously.¹⁸ 

The increased percentages of GFP-positive RBCs after the initiation of CID treatment preceded a delayed increase in total RBC number. For example, GFP-positive RBCs rose from approximately 5% at baseline to approximately 50% before the second dose of CID (Figure 4A). However, this dramatic rise in genetically modified RBCs was not reflected in an increase in the absolute number of RBCs measured at the same time (Figure 4D). We determined the absolute number of GFP-positive RBCs (Figure 4E) and the absolute number of GFP-negative RBCs (Figure 4F) in the treated and untreated mice. With the initiation of CID treatment, there was a clear increase in GFP-positive RBCs (Figure 4E) and a coincident decrease in GFP-negative RBCs (Figure 4F). This resulted in a large increase in the percentage of GFP-positive RBCs without a proportional increase in total RBC numbers. By increasing the frequency of CID administration, the total number of RBCs in the CID-treated mice rose to levels significantly higher than in the control mice (Figure 4D, G). We conjectured that the suppression of nontransduced RBCs after the initiation of CID treatment might have resulted from an inhibition of erythropoietin-dependent erythropoiesis. This interpretation was supported by an initial increase in total RBC numbers that was detectable at a single time point 3 weeks after the first dose of AP20187 (Figure 4D, asterisk) and was accompanied by a transient decrease in reticulocyte count (Figure 4C, asterisk). These findings suggested that AP20187 induced a surge in genetically modified RBCs and a temporary increase in RBC numbers, resulting in a physiologically relevant decrease in circulating erythropoietin levels, a decrease in reticulocyte count, and a sharp decrease in short-lived PK deficient RBC levels. We proceeded to directly test whether CID administration could suppress circulating levels of erythropoietin.

To determine whether CID treatment could reduce erythropoietin levels and, therefore, erythropoietin-dependent erythropoiesis, we measured erythropoietin levels in mice after CID administration. All the surviving mice (2 treated, 4 control) from the group shown in Figure 4 were observed for 12 weeks after the last course of CID, by which time the percentage of GFP-positive RBCs decreased to that at the previous baseline (Figure 5A). Three mice (2 treated, 1 control) were treated with CID for 4 weeks to increase the peripheral blood RBC numbers and the percentage of GFP-positive RBCs (Figure 5A). Mice were then killed, and erythropoietin levels were measured. As shown in Figure 5B, mice that received CID had erythropoietin levels that were intermediate between the levels measured in CBA/N and those measured in CBA-Pk-1^slc/Pk-1^slc mice, demonstrating that CID administration can suppress erythropoietin levels through the selective expansion of transduced erythrocytes.

The experiments outlined in Figures 4 and 5 demonstrate that erythropoiesis from normal donor cells can be selectively expanded using CIDs, resulting in CID-dependent increases in RBC number. These findings also indicate that CID treatment can suppress the erythropoietin-dependent generation of PK-deficient RBCs. We cannot exclude the involvement of an erythropoietin-independent process in which CID-responsive increases in GFP-positive RBCs result in the suppression of erythropoiesis arising from the nontransduced RBC compartment.

Further analysis of the 6 surviving animals 67 weeks after transplantation is presented as supplemental data (available at the Blood website; see the Supplemental Materials link at the top of the online article). However, conclusions related to toxicity are limited given the small numbers of experimental animals.

We also examined whether CID administration expanded transduced peripheral blood leukocytes. Because of the relatively large volume of blood required for these studies, leukocytes were evaluated only after a clear effect of CID treatment on RBCs was demonstrated. Retro-orbital blood sampling was performed on the mice presented in Figure 4 at week 44. DNA was purified from the leukocytes. PCR was performed using primers that flanked the mutation responsible for the enzyme deficiency. The mutation resulted in the loss of a BstEII restriction enzyme site, allowing the determination of donor leukocyte contribution (Figure 6A). Resultant bands were separated on an agarose gel, and band intensity was determined (Figure 6B). The average of the donor contribution in CID-treated and untreated mice is shown in the bar graph in Figure 6C. Donor contribution in untreated mice was 6.13%, and in treated mice it was 9.16%. This difference was not significant (P = .28). Leukocytes were also examined for GFP expression by flow cytometry. There was no significant difference in the proportion of neutrophils or lymphocytes that expressed GFP between the treated and untreated groups (Figure 6D). Unlike previous findings in the dog transplantation model,¹⁹  lymphocyte expansion was not observed.

A new approach to cell therapy aims to identify methods to specifically control the fate of transplanted cells.³³  Hemopoietic growth factors provide a means for controlling cell growth in vivo. However, growth factors lack the ability to selectively expand a subpopulation of genetically modified cells within a particular hemopoietic lineage. Selectivity can be achieved by engineering cells ex vivo so their fate can be controlled in vivo using exogenously supplied small molecules. We have developed a means whereby cell expansion can be directed using CIDs.¹⁶  This approach could be advantageous in gene therapy in which only a fraction of cells contain the therapeutic gene or for nonmyeloablative transplantation in which donor and host hemopoietic elements coexist.

In the studies outlined here, selective expansion of normal donor erythrocytes with normal pyruvate kinase activity resulted in improvements in the anemia of the recipient mice. CIDs functioned similarly to erythropoietin, targeting erythroid cells expressing the F36Vmpl fusion protein. Much like the use of growth factors, the dose of CID can be adjusted to achieve the desired level of RBC production (Figure 4E). The combination of reduced intensity conditioning followed by repeated administrations of CID was well tolerated by the CBA-Pk-1^slc/Pk-1^slc mice. One concern with repeated administrations of CIDs was that transduced progenitors would be deleted. Initial resistance to CID-mediated expansion in some mice was overcome with more frequent administration of CIDs. This resulted in a median RBC number in the normal range for the 5 treated mice. In fact, erythrocytosis developed in 2 of the CID-treated mice with RBC numbers exceeding 13 × 10⁹ RBCs/mL, necessitating a pause in CID treatment. We monitored mice that underwent transplantation for the development of therapy-associated leukemia, as has been reported in other murine gene transfer experiments.³⁴  Marrow from treated and untreated animals appeared morphologically indistinguishable, though one untreated mouse that underwent transplantation did acquire severe anemia 67 weeks after transplantation that was not associated with an increase in GFP-positive cells (Supplemental Materials).

Selectively expanding the numbers of transduced, normal donor RBCs allowed for the expansion of total RBC mass and a consequent suppression of erythropoietin. In this manner, the recipient, abnormal erythropoiesis was further suppressed. Figure 7 provides a schematic overview of the hematologic consequences of CID-mediated expansion of normal donor RBCs. Mice that underwent transplantation resulting in chimerism had 3 populations of marrow stem cells: (1) CBA-Pk-1^slc/Pk-1^slc recipient cells (PK-/-; open boxes); (2) nontransduced CBA/N donor cells (hatched); and (3) transduced, CID-responsive, CBA/N donor cells (black boxes). In the absence of CIDs, the donor RBCs had a survival advantage and increased relative to recipient cells. CID treatment expanded the transduced RBC population, resulting in a decrease in erythropoietin and a concomitant reduction in abnormal erythropoiesis, further enhancing selection of the normal donor cell population. Nevertheless, we cannot exclude the possibility of an erythropoietin-independent process in which the CID-responsive RBC population directly suppresses production from the nontransduced, CID-insensate, RBC compartment. One erythropoietin-independent process could be the marked splenomegaly that occurs during CID administration and could result in nonspecific RBC sequestration. In the setting of a fixed number of nontransduced RBCs and increasing numbers of GFP-positive RBCs, total numbers of GFP-negative RBCs would appear to decrease. We did not observe prolonged selection after CID withdrawal, arguing against an effect at the stem cell level. This result is in contrast to other selection strategies that rely on cytotoxic agents to select for stem cells that express chemoprotectant proteins from drug resistance genes.^32,35,36 

However, drug resistance genes have encountered 2 types of obstacles. First, stem cells and progenitors are relatively resistant to many cytotoxic drugs, resulting in relatively limited in vivo efficacy for drug resistance genes such as dihydrofolate reductase^37-39  and MDR1.^40,41  Second, the drugs used to achieve selection harbor considerable hematologic and nonhematologic toxicities.^39,42,43  In addition to immediate toxicities, the mutagenicity of many cytotoxic agents, most notably alkylating drugs used in association with methylguanine methyltransferase (MGMT), adds a significant burden of secondary malignancies, especially leukemias.^44,45 

Compared with selection using chemoprotectant proteins, CID-dependent expansion is reversible and does not require the systemic administration of DNA-damaging agents. However, limiting the proliferative response to the progeny of transduced stem cells would require continuous CID administration to maintain the therapeutic effect. Leukemia is generally thought to originate in cells with unlimited capacity for self-renewal, such as B cells, T cells, and hemopoietic stem cells.⁴⁶  Based on this principle, directing a growth signal to cells with a limitless capacity for self-renewal may raise a greater concern for the development of leukemia than if the growth signal were confined to committed progenitors, which are generally considered to have finite capacity for self-renewal. This view is supported by the abundance of clinical safety data that exist for chronically administered growth factors such as erythropoietin, granulocyte macrophage-colony-stimulating factor (GM-CSF) and G-CSF.^47,48 Although the requirement for ongoing CID administration constitutes an inconvenience and poses the risk for long-term drug exposure, the ability to withdraw selection may provide a crucial safety valve if leukemia were to occur. Testing for the development of leukemias in mice treated with CID may require performing secondary transplantation, similar to previous reports of leukemogenesis in mouse transplantation models.³⁴ 

This method could be applied as adjunct therapy to gene transfer-based therapies for erythrocyte disorders and for transplantation strategies aimed at establishing mixed chimerism. Improvements to this approach include the use of erythrocyte-specific promoters such as spectrin or ankyrin to limit expression of the selectable protein to erythroid progenitors.^49-51  In addition, we are testing other signaling molecules, including the J^H1 domain of Janus kinase 2 (Jak2), which may provide a more selective signal for directing RBC expansion (Shengming Zhao, M.W., Kenji Ihara, R.E.R., and C.A.B., manuscript submitted, January 2004). The combination of reduced-intensity conditioning with in vivo selection could prove useful for eventual gene therapy for the erythrocyte disorders.³² 

We thank James Yan for technical assistance.