Expansion of haematopoietic stem cells from normal donors and bone marrow failure patients by recombinant hoxb4


Neal S. Young, MD, Hematology Branch, National Heart, Lung, and Blood Institute, National Institutes of Health, Building 10 Clinical Research Center Room 3E-5140. 10 Center Drive, Bethesda, MD 20892-1202, USA. E-mail: youngn@nhlbi.nih.gov


In this study six versions of recombinant human hoxb4 proteins were produced and their effectiveness evaluated in expanding human haematopoietic stem and progenitor cells in vitro and in vivo. An N-terminal-tat and C-terminal histidine-tagged version of hoxb4 (T-hoxb4-H) showed the highest activity in expanding colony forming cells (CFCs) and long-term culture-initiating cells (LTC-ICs) when used at 50 nmol/l concentration in cell culture. Human cord blood CD34+ cells cultured with 50 nmol/l T-hoxb4-H showed a significant increase in severe-combined immunodeficient mouse-repopulating cells (SRCs). In a mouse model of immune-mediated bone marrow (BM) failure, T-hoxb4-H showed an additive effect with cyclosporine in alleviating pancytopenia. In addition, T-hoxb4-H expanded CFC and LTC-IC on BM samples from patients with refractory severe aplastic anaemia and myelodysplastic syndromes: after culturing with 50 nmol/l T-hoxb4-H for 4 d, BM cells from 10 of the 11 patients showed increases in CFC and LTC-IC, and the increase in LTC-IC was statistically significant in samples from four patients. Recombinant human hoxb4 could be a promising therapeutic agent for BM failure.

Deficiency in long-term haematopoietic stem cells (HSCs) and short-term progenitor cells is characteristic of human bone marrow (BM) failure syndromes, such as aplastic anaemia (AA) and hypocellular myelodysplastic syndromes (MDS), and leads to the clinical manifestation of severe BM aplasia and fatal pancytopenia (Young et al, 2006; Nakao, 2008). CD34+ cell number, and haematopoietic progenitor cell colony formation as reflected in quantitation of haematopoietic progenitor cells using the long-term culture initiating cell (LTC-IC) assay, are markedly decreased in AA and MDS (Maciejewski et al, 1994, 1996; Sato et al, 1998). Immunosuppressive therapy (IST) is effective in improving blood counts in 60–70% of AA patients. However, low blood counts often persist and relapse is frequent, requiring repeated treatment (Scheinberg et al, 2006). While growth factors have been used in addition to IST to support these patients, responses are usually limited to a single cell lineage, and in many cases, patients are not responsive to the treatment (Young & Maciejewski, 1997). Stem cell replacement therapy provides a definitive cure, but difficulties in obtaining well-matched donors and transplantation-associated complications limit this option to only a minority of AA patients (Young et al, 2006).

An agent that could expand patient residual HSCs would be useful in the treatment of AA and other BM failure syndromes. Many attempts have been made to expand HSCs in vitro using different combinations of cytokines, with disappointing outcomes despite maintenance of long-term repopulating stem cells can be achieved in the best circumstances (Conneally et al, 1997; Tisdale et al, 1998; Ueda et al, 2000; Gammaitoni et al, 2003). Thus, attention has shifted to transcription factors that govern stem and progenitor cell fate decisions. One well-studied factor is the homo-box B4 gene (HOXB4), a member of the homo-box family of transcription factors. Retroviral expression of HOXB4 in mice significantly improved HSC regeneration in vivo, with three log increases of HSCs in both primary and secondary recipients (Sauvageau et al, 1995; Thorsteinsdottir et al, 1999; Antonchuk et al, 2001), without affecting normal differentiation or inducing cell transformation (Sauvageau et al, 1995). Similarly, retroviral expressed hoxb4 expanded mouse HSCs by more than 1000-fold in vitro with the expanded HSCs remaining multipotent (B, T, and myeloid) and competitive in repopulating primary and secondary recipients (Antonchuk et al, 2002). In one report, retroviral-expressed human hoxb4 expanded human HSCs in vitro, but the effect was much less than that observed in the mouse (Buske et al, 2002). Recombinant hoxb4 has been produced in order to avoid the potential toxicities of retroviral vectors. A recombinant hoxb4, tat-hoxb4, expanded mouse HSCs (Krosl et al, 2003). However, the effect of recombinant human hoxb4 on HSCs from normal human donors and BM failure patients is unknown.

In this study, we produced six versions of recombinant human hoxb4 with the purification tag, six histidines, and the cell permealizaition tag, tat, at different locations relative to hoxb4, and tested their effects on human HSCs using colony-forming-cell (CFC) and LTC-IC assays. We selected one version of hoxb4 with the tat tag at the N-terminal and the histidine tag at the C-terminal (T-hoxb4-H) that had the highest HSC-expansion activity, determined its optimum working concentration, and tested its effectiveness in expanding severe-combined immunodeficient (SCID) mouse-repopulating cells (SRC) in human cord blood CD34+ cells. T-hoxb4-H was also tested in vivo in a mouse model of BM failure in conjunction with the immunosuppressive agent, cyclosporine. Further, the effectiveness of T-hoxb4-H was examined in ex vivo expansion of CFC and LTC-IC from normal volunteers, as well as from AA and MDS patients. Our results showed that recombinant hoxb4 expanded HSCs and could be a potential therapeutic agent for BM failure syndromes.

Materials and methods

Cloning, expression, and purification of recombinant human hoxb4

A commercial pET-21(+) vector (Novagen, Gibbstown, NJ, USA) was used to clone three expression vectors, Pet I, II and III, that expressed target proteins with tat and histidine tags at the C terminal (I), tat at the N-terminal and histidines at the C terminal (II), and both histidine and tat tags located at the N-terminal (III). The pET-21(+) vector DNA was digested by either XhoI or BamH1, and was dephosphorylated with calf intestinal phosphatase. Three primer sets (Pet-I: 5′- TCGAGTACGGACGCAAAAAGAGAAGGCAACGGCGACGTG (forward), 5′ TCGACACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTAC (reverse); Pet-II: 5′GATCATGTACGGACGCAAAAAGAGAAGGCAACGGCGACGTG (forward), 5′ GATCCACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTACAT (reverse); Pet-III: 5′ GATCATGCACCATCACCATCACCATTACGGACGCAAAAAGAGAAGGCAACGGCGACGTGGTGGGGGCGGAG (forward), 5′GATCCTCCGCCCCCACCACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTAATGGTGATGGTGATGGTGCAT (reverse) were melted at 95°C and annealed at gradually decreasing temperature to room temperature. The annealed primers were phosphorylated and ligated to either the XhoI or BamH1 site of linerized pET-21(+) vector in order to generate the pET-I, II and III vectors.

All versions of recombinant hoxb4 constructs were generated by DNA polymerase chain reaction (PCR) with primers (Version I: hox b4 tat his6: 5′CCACCTTGGATCCATGGCTATG AGTTCTTTTTTGATC (forward), 5′GATCAGCTCGAGCGCGCGGGGGCCTCCATTG (reverse); Version II: tat hoxb4 his6 5′CCACCTTGGATCCATGGCTATGAGTTCTTTTTTGATC (forward), 5′ GATCAGCTCGAGCGCGCGGGGGCCTCCATTG (reverse); Version III: his6-tat-hoxb4: 5′CCACCTTGGATCCATGGCTATGAGTTCTTTTTTGATC (forward), 5′GATCAGCTCGAGCTAGAGCGCGCGGGGGCCTCCATTG (reverse); Version IV: his6-hoxb4-tat-his6: 5′GATCATGCACCATCACCATCACCATTACGGACGCAAAAAGAGAAGGCAACGGCGACGTGGTGGGGGCGGAG (forward), 5′GATCCTCCGCCCCCACCACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTAATGGTGATGGTGATGGTGCAT (reverse); Version V: his6-tat-HA-hoxb4: 5′GATCATGCACCATCACCATCACCATGGTGGGGGCGGATACGGACGCAAAAAGAGAAGGCAACGGCGACGTGGTGGGGGCG (forward), 5′GATCCGCCCCCACCACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTATCCGCCCCCACCATGGTGATGGTGATGGTGCAT (reverse); Version VI: his6-hoxb4-HA-tat: 5′GATCATGCACCATCACCATCACCATGGTGGGGGCGGATACGGACGCAAAAAGAGAAGGCAACGGCGACGTGGTGGGGGCGGTTACCCATACGATGTTCCAGATTACGCTG (forward), 5′GATCCAGCGTAATCTGGAACATCGTATGGGTAACCGCCCCCACCACGTCGCCGTTGCCTTCTCTTTTTGCGTCCGTATCCGCCCCCACCATGGTGATGGTGATGGTGCAT (reverse) using a full-length HOXB4 cDNA clone (MHS1010-9205037, accession numberBC049204), purchased from Open Biosystems, Huntsville, AL, USA. Because of the high GC content in HOXB4 cDNA, Clontech advantage GC2 PCR kit (Clontech, Mountain View, CA, USA) was used for the PCR. Next, PCR fragments were purified and cloned into PCR2.1 vector (Invitrogen, Carlsbad, CA, USA). The desired fragments were subcloned further into pET- I, II and III vectors. All final clones were verified by sequencing and transformed into BL21 (DE3) cells (Novagen).

Transformed bacteria were grown in LB culture media (with ampicillin) to an optical density of 0·8 followed by induction with 1 mmol/l IPTG (Isopropyl β-D-1-thiogalactopyranoside) for 2–3 h. Bacteria were harvested, centrifuged (3000 g, 10 min at 4°C) and sonicated in binding buffer (6 mol/l GnHCl, 100 mmol/l NaH2PO4, 10 mmol/l Tris, pH 8·0). Cell lysates were centrifuged (20 000 g, 30 min at 4°C), resuspended in the binding buffer with 10 mmol/l imidazole, loaded on HisTrap (ni-affinity) chelating columns, and eluted with a 10–500 mmol/l imidazole gradient in binding buffer using AKTA fast protein liquid chromatography (FPLC) (Amersham Biosciences, Piscataway, NJ, USA). Hoxb4-containing fractions were desalted on a PD-10 column and eluted with ion-exchange buffer (10 mmol/l Tris, pH 8·0, 10 mmol/l NaCl). Next, the samples were loaded on HiTrap sepharose (SP) high performance (HP) (ion-exchange) column and eluted with a 10–500 mmol/l NaCl gradient in ion-exchange buffer on AKTA FPLC. High protein-containing fractions were again desalted and eluted in phosphate-buffered saline (Quality Biologicals, Gaithersberg, MD, USA). Endotoxin was removed by mixing samples with 1% Triton X-114 at 4°C for 10 min, transferred to a 37°C water bath for 10 min followed by centrifugation at 15 000 g for 10 min at room temperature, and the upper liquid phase was collected and subject to the above process two more times. Preparations were tested endotoxin-free by the Limuslus Amebocyte Lysate QCL-1000 (Cabrex Bioscience, Walkersville, MD, USA) endotoxin detection kit. The concentration of each protein was determined by first measuring total protein concentration using a bicinchoninic acid protein assay kit (Pierce, Rockford, IL USA), then multiplied by the proportion of the protein in the major band from all bands in the silver-stained polyacrylamide gel through image quantification using the ImageQuant TL software (Amersham Biosciences). The final products were supplemented with 5% glycerol, aliquoted and frozen at −80°C. All chromatography and desalting columns were purchased from Amersham Biosciences. All chemicals were purchased from Sigma-Aldrich, St Louis, MO, USA.

Sample collection

Human cord blood cells were kindly provided by the New York Blood Centre, NY. CD34+ cells were isolated using the indirect CD34 MicroBead Kit (Miltenyi Biotec, Bergish Gladbach, Germany). Heparinized human BM samples were obtained from the posterior iliac crest of patients and normal donors in five separate 0·5-ml aspirations. Cells were separated according to density differences by centrifugation on a Ficoll-hypaque gradient by the method of Boyum (1968). Cells present at the plasma/Ficoll-hypaque interface were collected and the red blood cells were lysed with ACK lysis buffer (Quality Biologicals, Gaithersburg, MD, USA). The cells were then washed in Hanks solution, frozen down in Dulbecco’s modified Eagle medium (DMEM) with 5% dimethyl sulfoxide and stored at −70°C using cryo-preservation boxes overnight before transferring to the liquid nitrogen tank the next day.

BM cells from nine patients with refractory AA and two patients with hypocellular MDS were used in the experiments; the cytogenetics of one patient were: inv 9 and t (7, 18), the other had marked morphologic dysplasia in both erythroid and myeloid cells with normal cytogenetics. Informed consent was acquired from each patient according to a protocol approved by the Institutional Review Board of the National Heart, Lung, and Blood Institute to collect patients’ samples for research studies. Samples were obtained in accordance with the Declaration of Helsinki for all patients.

Cell culture, CFC and LTC-IC assays

One to five ×105 BM cells were first cultured for 2 d in StemSpan medium (Stem Cell Technologies, Vancouver, BC, Canada) with 20 ng/ml recombinant human interleukin (IL)3, 20 ng/ml IL6, 100 ng/ml Flt-3, 100 ng/ml stem cell factor (SCF), 50 ng/ml thrombopoietin (TPO), 100 μg/ml penicillin and 100 μg/ml gentamycin in 12 well plates. Then recombinant hoxb4 protein and the control (bovine serum albumin, BSA) were added every 6 h at the desired concentrations. Cells were resuspended in fresh medium containing fresh cytokines and antibiotics every 24 h, and were harvested after culturing for four more days.

Human clonogenic progenitors were assayed by plating cells in 1 ml MethoCult with cytokines (Stem Cell Technologies) in 35 mm culture dishes for 15 d at 37°C with 5% CO2 and 95% humidity. CFCs were then scored by observation under light microscope. LTC-IC were measured by culturing human haematopoietic cells (1000 cells/cm2) on 80-Gy preirradiated M2-10B4 murine fibroblasts (American Type Culture Collection, Manassas, VA, USA) as feeder-layer cells (3 × 104 cells/cm2) in bulk (100 mm dish) or under limiting-dilution conditions (96-well plate) in MyeloCult 5100 containing 10−4 mol/l hydrocortisone (Stem Cell Technologies). Cell cultures were maintained for 5 weeks, with half of the medium replaced every week. Cells were harvested and assayed for CFC as LTC-IC output. Total LTC-IC number was obtained by multiplying the frequency of LTC-ICs, as determined in the secondary CFC assay, by the total number of cells present after the 5-week primary long-term co-culture.

SRC assay

Human cord blood CD34+ cells were cultured at 105 cells per well for 4 d with or without 50 nmol/l of the recombinant T-hoxb4-H protein. Cells were harvested and injected into sub-lethally irradiated (2.5 Gy) non-obese diabetic severe combined immunodeficient (NOD-SCID) mice (The Jackson Laboratory, Bar Harbor, ME, USA); each recipient received cells equivalent to 104 or 5 × 103 starting CD34+ cells together with 105 CD34 accessory cells. Recipients were euthanized 8 weeks after transplantation and BM nucleated cells were analysed for the presence of human CD45+ cells by flow cytometry. Mice were considered positive for human HSC engraftment when at least 0·1% CD45+ human cells were detected in the mouse BM cells. SRC frequency was calculated based on the Poission Distribution using equation Pi = e(−N) × (Ni/i!) at P0 (Chen et al, 1999). BM cells from positive mice were further analysed by flow cytometry using antibodies for human CD14, CD15, CD19 and CD34 antigens (BD Biosciences, San Jose, CA, USA).

Recombinant hoxb4 treatment in a mouse model of immune-mediated BM failure

Twelve C.B10-H2b/LilMcdJ (C.B10) mice (The Jackson Laboratory) each received 5 Gy sublethal irradiation followed by the infusion of 5 × 106 B6 lymph node (LN) cells. The mice were then divided into four groups of three animals each: (i) no treatment control, (ii) cyclosporine 50 μg/g/d i.p. once daily from days 0–6, (iii) cyclosporine 50 μg/g/d i.p. once daily from days 0–6 plus T-hoxb4-H 2 μg/g/d i.p. once daily from days 7–13, and (iv) T-hoxb4-H 2 μg/g/d i.p once daily from days 7–13. Animals were bled on weeks 2, 3, 5, 6, and 14 for complete blood counts. Animal survival was assessed at weeks 5, 6, and 14. All animal study protocols were approved by the Animal Care and Use Committee of the National Heart, Lung, and Blood Institute.

Statistical analysis

Data were analysed by the JMP statistical software on the fit-model platform (SAS Institute Inc., Cary, NC, USA). Results are presented as means with standard errors. Statistical significances were declared at P < 0·05 and P < 0·01 respectively.


Expression and purification of recombinant human hoxb4

We expressed six versions of recombinant human hoxb4 with the tat and histidine tags at different locations (Fig 1A). Version V is the same as described by Krosl et al (2003). Proteins were purified by nickel-affinity chromatography followed by ion-exchange chromatography. Version III could not be purified by nickel-affinity chromatography, possibly because the histidine residues were not accessible to the nickel-affinity beads. After purification, endotoxin was removed using Triton X-114 extraction. Purified hoxb4 proteins were visualized by silver staining after polyacrylamide gel electrophoresis (PAGE) (Fig 1B).

Figure 1.

 Expression of recombinant human hoxb4. (A) Diagram of six versions of recombinant hoxb4 with different locations of the histidine and tat tags. (B) Escherichia coli expressed, nickel-affinity chromatography and HiTrap SP HP ion-exchange chromatography purified hoxb4 proteins were analysed by SDS-PAGE and stained with Silver Quest staining kit from Invitrogen. Lanes 1, 2, 3 and 4: BSA 1 μg, 0·2 μg, 0·1 μg and 0·02 μg; lanes 5, 6, 7, 8 and 9: version I, II, IV, V and VI hoxb4 each 0·6 μg (3 μl) before endotoxin removal; lanes 10, 11, 12, 13 and 14: version I, II, IV, V and VI hoxb4 each 3 μl after endotoxin removal.

Expansion of CFC and LTC-IC by recombinant human hoxb4

The effects of the five versions of purified hoxb4 protein were tested on fresh human BM cells. Normal BM mononuclear cells were cultured in StemSpan medium supplemented with recombinant human IL3, IL6, Flt-3, SCF and TPO for 2 d at 5 × 105 cells/well. Different versions of hoxb4 or control (BSA) were added at 10 nmol/l or 50 nmol/l respectively every 6 h for 4 d, similar to the published conditions (Krosl et al, 2003). Cells were then collected and assayed for CFC and LTC-IC. Version II hoxb4 (T-boxb4-H) at 10 nmol/l and versions I, II, IV and V at 50 nmol/l all induced significant CFC expansion (P < 0·05, Fig 2A), and versions I, II, V and VI at 50 nmol/l showed significant expansion of LTC-IC (P < 0·05, Fig 2B) compared with the control (BSA). T-hoxb4-H demonstrated the largest expansion in CFC (9·6 ± 0·7-fold, Fig 2A) and LTCIC (14·2 ± 1·0-fold, Fig 2B). We further determined an optimum concentration for T-hoxb4-H by culturing 5 × 105/well frozen normal BM cells with 0, 10, 50, 100, 150, 200 and 300 nmol/l T-hoxb4-H, with the protein added every 6 h. After 4 d in culture, cells were harvested and assayed for CFC. The expansion of CFC reached a peak at 50 nmol/l (Fig 2C). Therefore, 50 nmol/l of T-hoxb4-H was chosen for further experiments.

Figure 2.

 Expansion of HSCs by recombinant human hoxb4. 1–5 × 105 normal human BM cells were cultured in 1 ml cytokine-supplemented StemSpan medium for 2 d, then recombinant hoxb4 proteins were added at desired concentrations once every 6 h for 4 more days. Cells were harvested, 10% of the cells were used for CFC assay and 90% of the cells were used for LTC-IC assay. (A) Overall, hoxb4 had a significant effect (P < 0·01) in CFC expansion, of which version II at 10 nmol/l concentration and versions I, II, IV and V at 50 nmol/l concentrations significantly expanded CFC (P < 0·05) compared with cultures without hoxb4. (B) Hoxb4 also had a significant effect (P < 0·01) in expanding LTC-IC, of which all hoxb4 versions at 50 nmol/l concentration significantly expanded LTC-IC compared with the controls. (C) Among all five versions of hoxb4, version II had the highest activity in expanding fresh human BM cells. Dose response of version II hoxb4 (T-hoxb4-H) was tested in frozen human BM cells. T-hoxb4-H used at 50, 100, and 150 nmol/l concentrations showed significantly higher (P < 0·05) CFC expansion than controls.

Expansion of human cord blood SRC by recombinant hoxb4

Next, we cultured human cord blood CD34+ cells with or without 50 nmol/l T-hoxb4-H for 4 d. The cells were harvested and injected into sublethally irradiated NOD/SCID mice, with each recipient receiving cells equivalent to 104 or 5 × 103 starting CD34+ cells. Eight weeks post cell transplantation, BM cells were harvested from the recipients and subjected to flow cytometric analysis with an anti-human CD45 antibody. Based on the number of recipients that did not have at least 0·1% of human CD45+ cells in the BM, we calculated SRC concentration using the Poisson probability at P0. In three separate experiments, cord blood CD34+ cells cultured with T-hoxb4-H had significantly higher (P < 0·01) SRC frequencies than did cells cultured without T-hoxb4-H (Table I). The engrafted cells showed similar myeloid and lymphoid differentiation in hoxb4 treated and control groups as analysed by flow cytometry with myeloid and lymphoid markers (Fig 3).

Table I.   Expansion of SRC by recombinant hoxb4.
ExperimentTreatmentPositive/ total% PositiveSRC frequency
  1. Human cord blood CD34+ cells were cultured for 4 d in StemSpam media with 50 nmol/l BSA control or T-hoxb4-H. Cells were harvested, washed and injected into sub-lethally-irradiated (2.5 Gy) NOD/SCID mice, each mouse received cells equivalent to 104 (experiment I) or 5 × 103 (experiment II and III) starting CD34+ cells. Animals were euthanized 8 weeks later; the presence of human CD45+ cells in the recipient BM cells was analysed by flow cytometry. Mice with 0·1% or higher human CD45+ cell in the BM were recorded as positive of human HSC engraftment. SRC frequency was calculated by using the Poission equation Pi = e(−N) × (Ni/i!) to calculate P0. Data shown were from three independent experiments.

IControl9/15601 in 10989
IT-hoxb4-H13/17761 in 6910
IIControl4/15261 in 16118
IIT-hoxb4-H7/14501 in 7213
IIIControl4/15261 in 16118
IIIT-hoxb4-H8/14571 in 5901
Figure 3.

 Multilineage engraftment of hoxb4-expanded human cord blood CD34+ cells. Representative flow cytometry profile of engrafted BM cells in mice received BSA or T-hoxb4-H treated human cord blood CD34+ cells. Human cord blood CD34+ cells were cultured for 4 d with 50 nmol/l BSA or T-hoxb4-H. Cells were then collected and injected into sublethally irradiated NOD/SCID mice, with each recipient receiving cells equivalent to 104 or 5 × 103 starting CD34+ cells in a SRC assay as detailed in Materials and methods. Recipient BM cells were harvested 8 weeks later and stained for human myeloid (CD14/CD15) and lymphoid (CD19) markers.

Effect of human hoxb4 in a mouse model of immune-mediated BM failure

Human hoxb4 shares 96% protein sequence homology with mouse hoxb4, thus, we investigated whether the recombinant T-hoxb4-H protein had a protective role in a mouse model of immune-mediated BM failure. As previously reported, infusion of 5 × 106 C57BL/6 (B6) LN cells to minor-histocompatibility antigen-mismatched C.B10 mice leads to rapid development of BM failure (Chen et al, 2007). These mice were treated with either cyclosporine 50 μg/g/d, T-hoxb4-H 2 μg/g/d (based on the optimum 50 nmol/l concentration for cell culture), or both and performed serial complete blood counts (CBC). Mice treated with both cyclosporine and T-hoxb4-H had higher blood cell counts, especially white blood cell and neutrophil counts, than did mice treated with cyclosporine alone, although the differences were not large enough to be statistically significant (P > 0·05, Fig 4A). Although T-hoxb4-H treatment alone did not protect the mice from BM failure, one mouse that received T-hoxb4-H survived and recovered, while all mice in the control group died within 5 weeks due to BM failure (Fig 4B). Animals that survived to week 6 all survived to week 14 with normal blood counts. Recombinant human hoxb4 exerted a positive effect in alleviating pancytopenia in mice with immune-mediated BM failure.

Figure 4.

 Hoxb4 enhances cyclosporine in rescuing animals in a mouse model of immune-mediated BM failure. Twelve C.B10 mice each received 5 Gy TBI and infused with 5 × 106 B6 LN cells and divided into four groups of three mice each: (i) No treatment control, (ii) Cyclosporine 50 μg/g/d i.p. once per day for 7 d from day 0 to 6, (iii) Cyclosporine 50 μg/g/d i.p. once per day for 7 d from day 0 to 6 plus hoxb4 at 2 μg/g/d i.p. once per day for 7 d from day 7 to 13, and (iv) hoxb4 at 2 μg/g/d i.p once per day for 7 d from day 7 to 13. (A) Complete blood counts were analysed weekly 2 weeks after LN cell infusions. (B) Animal survival was recorded at week 5 and 6. Animals that survived to week 6 all survived to week 14, when all animals were euthanized.

Effect of recombinant human hoxb4 on BM cells from patients with BM failure

To examine the effect of recombinant hoxb4 on patient BM cells, we cultured frozen BM cells from patients with refractory AA (n = 9) or MDS (n = 2) in comparison to a normal control (Table II) with or without the addition of the 50 nmol/l T-hoxb4-H protein. After 4 d in culture, T-hoxb4-H-treated cells showed higher CFCs in 10 out of 11 patient samples as well as in the normal control sample, with the overall hoxb4 treatment effect being statistically significant (P < 0·01, Fig 5A). Treatment with T-hoxbH resulted in a lower average CFC expansion in samples from AA patients (1·8 ± 0·2-fold) than in samples from MDS patients (2·6 ± 0·5-fold) and the normal control (2·4 ± 0·7-fold) (P < 0·09, Fig 5B). T-hoxb4-H treatment also increased LTC-IC in 10 out of 11 patient samples, with the overall effect being statistically significant (P < 0·01, Fig 5C). Four of these 10 samples showed a significant (P < 0·05) increase in LTC-IC by T-hoxb4-H treatment in comparison to the same four samples cultured with BSA. Again, T-hoxb4-H-induced LTC-IC expansion was lower in samples from AA patients (3·8 ± 0·5-fold) than in samples from MDS patients (12·2 ± 1·1-fold) and the normal control (17·4 ± 1·6-fold) (P < 0·0001, Fig 5D).

Table II.   Patient information.
Patient no.Date of sample collectionCell numberDiagnosis
  1. Summary of information on 11 patients whose BM cells were used in the expansion experiment. Both myelodysplatic syndrome (MDS) patients had refractory cytopenia with multilineage dysplasia (RCMD). AA, aplastic anaemia.

Control08/20051·6 × 105Normal
00102/14/20063·0 × 105AA
00202/14/20062·6 × 105AA
00302/07/20061·4 × 105AA
00407/12/20052·6 × 105AA
00502/07/20062·4 × 105AA
00601/24/20061·0 × 105AA
00701/24/20065·5 × 105AA
00805/10/20050·4 × 105AA
00909/27/20052·0 × 105AA
01004/20/20040·2 × 105MDS
01106/01/20044·0 × 105MDS
Figure 5.

 Expansion of HSCs from BM failure patients by recombinant hoxb4. Frozen BM cells (0·2–5·5 × 105) from a normal control and BM failure patients were cultured in 1 ml cytokine-supplemented Stemspam medium for 2 d, then T-hoxb4-H protein was added at 50 nmol/l every 6 h for 4 d. Harvested cells were tested for the expansion of CFC and LTC-IC. (A) Treatment with hoxb4 significantly expanded CFC (P < 0·01) in normal control and patient samples, (B) samples from MDS patients showed higher CFC expansion than from AA patients although not statistically significant (P < 0·09). (C) Similarly, hoxb4 treatment produced significant LTC-IC expansion (P < 0·01) in samples from normal control and patients, and (D) samples from the two MDS patients showed significantly higher LTC-IC than samples from AA patients (P < 0·0001).


Primitive HSCs are the ultimate source of haematopoietic progenitor cells that sustain lifelong production of mature blood cells. In human BM failure diseases, such as AA and hypocellular MDS, immune attack or intrinsic cellular defects cause a severe HSC deficiency, resulting in marrow hypoplasia and pancytopenia, which is fatal if not treated (Maciejewski et al, 1994, 1996; Yamaguchi et al, 2003). Counteracting autoimmunity and providing normal functional HSCs are used to treat patients with BM failure.

Our current study aimed to explore a novel treatment strategy for BM failure syndromes by expanding residual HSCs using recombinant human hoxb4, a transcription factor that has been extensively studied for its role in promoting HSC self-renewal (Sauvageau et al, 1995; Antonchuk et al, 2002; Amsellem et al, 2003; Krosl et al, 2003). The choice of hoxb4 was based on previous studies in mouse models, in which the transduction of HSCs with a hoxb4-containing retroviral vector resulted in marked HSC expansion (Antonchuk et al, 2002; Sorrentino, 2004). In a mouse HSC transplantation model, lethally-irradiated recipients reconstituted with BM cells that had been transduced with a control vector had only 5–10% of normal number HSCs regenerated, whereas recipients reconstituted with BM cells transduced with a retroviral vector expressing hoxb4 had 100% of normal number of HSCs regenerated (Sauvageau et al, 1995). Expansion of HSCs did not continue after normal numbers of HSCs were regenerated, suggesting the effects of hoxb4 were subjected to normal homeostatic controls. When mouse HSCs were transduced with hoxb4 and cultured in vitro for 14 d, the number of HSCs expanded 40-fold from the initial numbers, or 2000-fold relative to control cultures with cytokines only (Antonchuk et al, 2002).

The mechanism of hoxb4-mediated HSC expansion is not well understood. Gene expression profiling in combination with subsequent functional analysis with enriched adult murine HSCs has suggested that hoxb4 modulated the response of HSCs to multiple extrinsic signals in a concerted manner, thereby shifting the balance toward stem cell self-renewal (Schiedlmeier et al, 2007). Hoxb4 might protect adult HSCs from detrimental effects mediated by the pro-inflammatory cytokine TNF-α. This protection also probably contributes to the competitive repopulation advantage of hoxb4-expressing HSCs observed in vivo (Schiedlmeier et al, 2007).

The high level of hoxb4 expression achieved with viral vectors provided the possibility of using hoxb4 for HSC expansion ex vivo for clinical use. However, one major concern is that uncontrolled high levels of hoxb4 expression, achieved by using viral vectors with strong promoters, have been associated with leukaemias in animal models, and abnormalities in myeloid differentiation in cell cultures (Schiedlmeier et al, 2003; Brun et al, 2004; Larochelle & Dunbar, 2008; Zhang et al, 2008). To avoid the use of viral vectors, recombinant hoxb4 fusion protein that contains a plasma-membrane permeabilization sequence ‘tat’ to allow the entrance of hoxb4 protein into cell cytoplasm was shown to stimulate HSC-expansion (Krosl et al, 2003).

We reasoned that the position of the purification and permeabilization tags could affect hoxb4 activity. Therefore, we generated six versions of recombinant hoxb4, varying in the locations of the histidine and tat tags. Indeed, one version of recombinant hoxb4 could not be purified. Of the five versions of purified human hoxb4, including version V in which the histindine and tat tag at the N-terminal of the protein was identical with the one already reported (Krosl et al, 2003), version II that had an N-terminal tat and a c-terminal histidines, showed the highest activity in expanding CFC (10-fold) and LTC-IC (15-fold) in fresh human BM cells in a 4-d culture at 50 nmol/l concentration. In a study using lentiviral HOXB4-transduced feeder layers, passively diffused hoxb4 induced about 20-fold expansion of LTC-IC and 2·5-fold expansion of SRC in human cord blood cells when cells were cultured for 5 weeks (Amsellem et al, 2003); we achieved comparable levels of LTC-IC and SRC expansion on fresh human BM cells and human cord blood CD34+ cells by culturing with the version II T-hoxb4-H protein in 4 d. In our experiments, there were no significant differences in total cell numbers between T-hoxb4-H- and BSA-treated fresh human BM cells, cord blood CD34+ cells, or patient BM cells, yet T-hoxb4-H orchestrated significant CFC, LTC-IC and SRC expansion, with the expansion of LTC-IC being much greater than that of CFC. These data are consistent with the previous report that hoxb4 exerts its activity on more primitive HSCs (Sauvageau et al, 1995).

Adding to the previously published data in murine models of marrow failure, in which IST with anti-thymocyte globulin, cyclosporine, or regulatory T cells ameliorated BM destruction (Bloom et al, 2004; Chen et al, 2007), recombinant hoxb4 demonstrated additional protection in conjunction with cyclosporine in preventing BM failure, possibly by expanding HSCs in vivo (It was not surprising the HSC expansion could not be achieved in this situation with hoxb4 alone without immunosuppression, as any gain in HSCs would be eliminated by active immune attack).

Using frozen BM cells, we found that recombinant hoxb4 expanded HSCs ex vivo from patients with AA and MDS. These patients were not candidates for haematopoietic stem cell transplant and had been refractory to IST. They could benefit from treatment to expand residual HSCs. Although the number of patient samples tested was small, the in vitro results are promising, because 90% of the samples responded positively to hoxb4 treatment, as assayed by CFC and LTC-IC. Cells from both MDS patients responded favourably to hoxb4 stimulation, with significant LTC-IC expansion, and cells from three of nine AA patients also showed significant LTC-IC expansion. Together with in vivo data from the mouse model, we hypothesize that hoxb4 might be useful in treating BM failure patients when combined with IST, and patients with higher number of residual HSCs would be more likely to benefit.

Conflict of interest disclosure

The authors declare no competing financial interests.


We thank Cynthia Dunbar and Andre LaRochelle from Haematology Brach, NHLBI for helpful discussions; Jan Melenhorst from Haematology Branch, NHLBI for providing human cord blood cells; and William DeGraff from Radiation Biology Branch, NCI for providing assistance in feeder layer cell irradiation. This research was supported by the Intramural Research Program of the National Heart, Lung and Blood Institute.