Molecular interactions involved in HOXB4-induced activation of HSC self-renewal

Most mature blood cells have short life spans, and are continuously generated from totipotent hemopoietic stem cells (HSCs). The distinct feature of HSCs is their ability to undergo self-renewal divisions, and to generate progeny that differentiate into mature cells of myeloid and lymphoid lineages. Although the genetic and molecular mechanisms responsible for the control of self-renewal and differentiation outcome of HSC divisions remain to be elucidated, it is becoming evident that Hox genes are involved in regulation of normal and leukemic hemopoiesis.^1,2 

Specifically, there is growing body of data demonstrating that in the mouse bone marrow (BM) transplantation model, retroviral overexpression of HOXB4 enhances HSC self-renewal and induces up to 1000-fold in vivo expansion.^3-5  Recent studies have also shown that HOXB4 induced rapid ex vivo expansion of the transduced HSCs⁶  and demonstrated that ex vivo HSC expansion is achievable using a soluble recombinant HOXB4 protein.^7,8  Importantly, the HOXB4-expanded HSCs retained their normal differentiation and long-term repopulation potential, and no hematologic abnormalities have been detected in large groups of mice that received transplants with HOXB4-transduced HSCs. In one study however, differentiation of human CD34⁺ cells was perturbed by the high expression levels of a fusion HOXB4 protein.⁹  Except for this study, the accumulated results indicate that HOXB4 activates a distinct pathway or pathways that enhances self-renewal divisions of HSCs without overriding the regulatory mechanisms that maintain normal steady-state hemopoiesis.

Molecular mechanisms and target genes responsible for HOXB4-induced HSC expansion remain to be elucidated. Due to the high degree of conservation within the homeodomain, different Hox proteins recognize similar regulatory sequences in vitro,¹⁰  although their specificity of sequence recognition,^11,12  affinity of DNA binding,^13,14  and transactivating potential^15-17  can be modified through heterodimerization with PBX and Meis/Prep members of the Three Amino acid Loop Extension (TALE) family of transcription factors. Hox proteins encoded by paralogs one to 10 cooperatively bind DNA with Pbx proteins through a conserved hexapeptide sequence (YPWM) preceding the homeodomain (HD) of Hox proteins.¹¹  Hox 9 through 13 proteins directly interact with Meis/PREP proteins.¹⁸  Paralogs 9 and 10 encode distinct proteins that interact with both PBX and PREP, and functional trimeric interactions between Hox, Pbx, and Meis on native Hox enhancers have been reported.^19,20  Genetic studies revealed the requirement for Hox/Pbx^16,21,22  and Hox/Pbx/Meis interactions^20,23,24  for execution of some developmental programs, whereas some functions of Drosophila Hox homologs Ubx and AbdA were shown to be Pbx independent.^25,26  Requirement for Hox/Pbx collaboration in regulation of hemopoietic cell behavior appears to be Hox protein– and cellular context–dependent. Hoxa9/Pbx interactions were essential for some²⁷  but not all aspects of Hoxa9-mediated immortalization of primary BM cells.²⁸  Mutations in the PBX interaction motif of Hoxb8 abrogated its ability to prevent differentiation of myeloid cells,²⁹  but not collaboration between Pbx1b and Hoxb3 or Hoxa9 in induction of leukemic transformation could be detected.^30,31  An additional level of complexity regarding the homeodomain protein activity was uncovered with Ftz whose activity in tissue patterning can be accomplished by a mutant lacking its HD.³² 

To further characterize the molecular mechanisms underlying the HOXB4-induced HSC self-renewal, we analyzed the requirement of the HOXB4-DNA and HOXB4-PBX interaction in execution of this function, by examining the in vitro and in vivo proliferation and differentiation capacity of primary BM cells engineered to express HOXB4 proteins carrying inactivation amino acid substitutions in the DNA-binding domain, or in the surface required for HOX-PBX interactions. We demonstrate that DNA binding is essential for HOXB4-induced proliferation of BM cells in vitro, and expansion of the transduced HSC population in vivo. We also show that direct HOXB4-PBX interactions involving the YPWM motif of HOXB4 are not required for its HSC-expanding activity, and suggest that the total levels of HOXB4 protein determine the magnitude of HOXB4-induced proliferation of primitive BM cells.

Bone marrow donors were (PepC3) (C57B1/6Ly-Pep3b × C3H/HeJ)F1 mice, and recipients were (C57B1/6J × C3H/HeJ)F1 (LY5.2⁺) mice and donor/recipients (C57B16/Pep3b) (Ly5.1⁺) and (C57B1/6J) (LY5.2⁺). All animals were housed and handled in accordance with the guidelines of the Clinical Research Institute of Montreal animal care and use committee, which also approved the experiments.

The HOXB4(W→G) mutation (Figure 1A row 3) was generated by replacing the 24-bp AccI–PmlI fragment of HOXB4 cDNA with the corresponding oligonucleotide containing a T⁴⁷²G substitution, and was verified by sequencing. Generation of HOXB4(N→A) mutant (Figure 1A row 2) was described previously.³³ 

To generate expression vectors encoding C-terminal FLAG-tagged proteins, wild-type (WT) and mutated forms of HOXB4 were subcloned upstream and in frame of FLAG epitope in pCMV-Tag (Stratagene, La Jolla, CA). In vitro transcription vectors were created by subcloning FLAG-tagged WT and mutated HOXB4 cDNAs into pcDNA3 (Invitrogen, Carlsbad, CA). Retroviral vectors encoding the FLAG-tagged WT and HOXB4(W→G) mutant were generated by subcloning the corresponding cDNA into myeloproliferative sarcoma virus phosphoglycerate kinase and/or neomycin resistance gene (MSCV-PGKneo^r) upstream of the PGKneo^r cassette (MSCV-HOXB4-FLAG-PGK-neo no. 830, MSCV-HOXB4(W→G)-FLAG-PGK-neo no. 867). MSCV–internal ribosome entry site (IRES)–green fluorescent protein (GFP) retroviral vectors encoding WT and both mutated HOXB4 forms were created by subcloning the various HOXB4 cDNAs upstream of IRES (MSCV-HOXB4-IRES-GFP no. 812, MSCV-HOXB4(N→A)-IRES-GFP no. 815 and MSCV-HOXB4(W→G)-IRES-GFP no. 827).

The retroviral vectors MSCV-PGK-neo no. 225, MSCV-HOXB4-PGK-neo no. 178, and MSCV-HOXB4(N→A)-PGK-neo no. 428 have been described before.^3,33 

Maintenance of ecotropic retrovirus packaging BOSC-23 and GP+E-86, and amphotropic vesicular stomatitis virus (VSV) pseudotyped retrovirus packaging cells 293 GPG, generation of high-titer helper-free GP+E-86 producer cells, and infection of primary BM cells were described.³⁴  BaF/3 cells were maintained and infected as described.³⁵ 

To exclude potential pseudo-transduction effect, BM cells recovered after coculture with retroviral producers were cultured for 2 days in Iscove modified Dulbecco medium (IMDM) supplemented with 15% fetal calf serum (FCS) and 10 ng/mL interleukin 3 (IL-3), then the transduced GFP⁺(HOXB4⁺) cells were sorted. Competitor cells were prepared as described.³⁶  Following a 1-day recovery period, liquid cultures were initiated by resuspending 5 × 10⁴ GFP⁺ BM cells/mL in IMDM with 15% FCS and 10 ng/mL IL-3. After indicated periods of growth, the viable (trypan blue negative) cells were counted and diluted with fresh media so that cell density was maintained between 5 × 10⁴ and 5 × 10⁵ cells/mL. At the same points in time, suitable aliquots of cultures were plated in methylcellulose containing 10 ng/mL IL-3, 10 ng/mL IL-6, 50 ng/mL stem cell factor (SCF), and 5 U/mL of erythropoietin (Epo). Colonies were scored on day 10. To determine the in vitro competitive proliferation potential of the transduced cells, cultures comprising 10% GFP⁺ plus 90% nontransduced competitors were initiated at a density of 5 × 10⁴ cells/mL, and the relative contents of GFP⁺ cells after 5- to 6- and 10- to 11-day incubations were determined by flow cytometry. To compare the in vitro competitive proliferation potential of HOXB4^high and HOXB4^low BM cells, liquid culture comprising 20% HOXB4^high (Ly5.1⁺ GFP^high) and 80% HOXB4^low (Ly5.2⁺ GFP^low) was initiated at 1 × 10⁵ cell/mL in IMDM supplemented with 15% FCS, 6 ng/mL IL-3, 10 ng/mL IL-6, and 100 ng/mL SCF, and proportions of HOXB4^high (Ly5.1⁺) cells were determined by flow cytometry at indicated time points. IL-3–induced proliferation responses of the transduced BM populations were determined using a ³H-thymidine incorporation assay as described.³⁷  Methylcellulose and CV-1 cell line transformed with an origin-defective SV40 virus (COS) cell supernatant-derived cytokines used for these experiments were prepared and quantitated at Institut de Recherches Cliniques de Montréal (IRCM). All other media components were purchased from GIBCO/Invitrogen (Burlington, ON, Canada).

Recipient mice were irradiated with 850 cGy (160 cGy/min, ¹³⁷Cs γ; J.L., Shepherd and Associates, San Fernando, CA). Groups of control neo^r, HOXB4(N→A)neo^r and HOXB4(WT)neo^r BM transplantation chimeras were generated by transplanting 1.2-1.5 × 10⁵ BM cells recovered from cocultures with retroviral producers as described.⁵  To generate groups of control GFP⁺, or HOXB4(W→G)GFP⁺ or HOXB4(WT)GFP⁺ recipients, BM cells recovered from cocultures with retroviral producers were cultured for an additional 3 days, then transplantation inocula (4 × 10⁵ cells/recipient) comprising 10% GFP⁺ cells were prepared by diluting the infected BM populations with nontransduced competitors. The frequencies of myeloid and lymphoid progenitors in the hemopoietic tissues of mice undergoing transplantation were determined as described.³⁸ 

HSC numbers in primary recipients were evaluated using a limiting dilution transplantation-based assay (competititive repopulation unit [CRU] assay) that detects cells with competitive, long-term lympho-myeloid repopulation capacity.³⁹  Primary mice were killed 12 weeks after transplantation, and their BM cells were transplanted into a series of secondary recipients at varying dilutions (5 × 10³ to 1 × 10⁶ cells/recipient, 5-10 recipients/dilution) along with 1 × 10⁵ freshly isolated helper bone marrow cells. The level of lymphoid and myeloid repopulation with Ly5.1⁺ or GFP⁺ (donor derived) cells in secondary recipients was determined 16 weeks after transplantation by flow cytometric analysis of peripheral blood.⁴⁰  CRU frequencies were calculated by applying Poisson statistics to the proportion of negative recipients at different dilutions using Limit Dilution Analysis software (Stem Cell Technologies, Vancouver, BC, Canada).

GFP⁺ cells were sorted using FACS^StarPlus equipped with a 488-nm argon laser or MoFlo (Cytomation, Fort Collins, CO). Proportions of Ly5.1⁺ cells were determined by flow cytometry at indicated days using phycoerythrin-conjugated Ly5.1 antibody (Pharmingen, Mississauga, ON, Canada). Discrimination by flow cytometry of myeloid and lymphoid cells was confirmed using allophycocyanin (APC)–conjugated antibodies recognizing cell surface lineage-specific markers (Mac-1, B220, CD4, and CD8; Pharmingen) as described.³⁶ 

Southern blot analyses were performed as described previously.³⁰  The probes used were 0.9-kb neo^r cDNA, 0.73-kb GFP cDNA, and 1.4-kb erythropoietin receptor cDNAs,³⁵  labeled with ³²P by random primer extension.

Total cell lysate and Western blot were performed as described.⁴¹  The primary antibodies used were mouse anti-FLAG antibody (Stratagene), rat anti-HOXB4 antibody (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City), mouse anti–β-tubulin (Sigma, St Louis, MO), and rabbit anti–c-Jun (Santa Cruz Biotechnology, Santa Cruz, CA) followed by horseradish peroxydase–conjugated antimouse, antirat, and antirabbit antibodies (Santa Cruz Biotechnology).

Pulse-chase experiments were performed as described.⁴²  The amount of radioactive and the total amount of FLAG-tagged proteins in each lane were measured using the STORM 860 and the ImageQuant 5.0 program (Molecular Dynamics, Sunnyvale, CA). The half-life was calculated from the proportions of radioactive proteins at the indicated time points using AllFit (Charles and André DeLéan, University of Montreal, QC, Canada).

EMSAs were performed as described previously¹³  using proteins produced by transcription and translation coupled reticulocyte lysate system (TNT) coupled as recommended in manufacturer's instructions (Promega, Madison, WI). The DNA probe was gel purified oligonucleotide 5′-CGA ATT GAT TGA TGC ACT AAT TGG AG-3′. For supershift analyses, DNA binding reactions were incubated for an additional 30 minutes with 1 μg anti-FLAG antibody (Stratagene).

The requirement for HOXB4 DNA binding in regulation of HOXB4-induced expansion of HSCs was examined using point mutation N⁵¹→A within helix 3 of the homeodomain (Figure 1A row 2, HOXB4(N→A)),^43,44  while the role of direct HOXB4-PBX interactions was assessed through W149G mutation in the conserved tetrapeptide (Figure 1A row 3, HOXB4(W→G)). ⁴⁵  Wild-type (HOXB4(WT)) and mutant HOXB4 cDNAs produced proteins of the predicted sizes as detected by Western blot analyses of the in vitro–transcribed and translated protein carrying a C-terminal FLAG epitope (Figure 1B). EMSAs confirmed that the N⁵¹A mutation in the homeodomain of HOXB4(N→A) abolished the DNA binding capacity of protein (Figure 1C, compare lanes 3-5 with lanes 6-8). HOXB4(W→G) retained the ability to bind DNA as monomer (Figure 1C, lane 9 vs lane 3) but could not form heterodimers with PBX1 (Figure 1C, lanes 3-5 vs lanes 9-11).

HOXB4 overexpression confers a pronounced in vitro proliferation advantage to transduced BM cells.^5,6  To determine whether this HOXB4 function depends on its ability to bind DNA, mouse BM cells were infected with recombinant retroviruses encoding HOXB4(WT) and GFP, or HOXB4(N→A) and GFP, or GFP alone. In the in vitro competitive proliferation assays initiated with 10% of GFP⁺ cells (HOXB4(WT)⁺, HOXB4(N→A)⁺, or control GFP⁺), the proportions of HOXB4(WT) cells increased to approximately 80% within 10 days (Figure 2A). In cultures initiated with HOXB4(N→A), or with control GFP⁺ cells, no noticeable increase in the proportions of GFP⁺ cells above the starting 10% could be detected. The in vitro expansion of myeloid clonogenic progenitors (CFCs) was determined in cultures initiated with equal numbers of GFP⁺ cells sorted from the 3 populations examined. During a 10-day expansion, HOXB4(WT) CFCs increased more than 1000-fold over initial numbers, compared with a 5- to 8-fold increase determined for HOXB4(N→A) and control GFP CFCs (Figure 2B), suggesting that DNA-binding activity of HOXB4 is needed to enhance the in vitro proliferation capacity of the transduced BM cells.

The correlation between the in vivo HSC expansion of HOXB4 and its ability to bind DNA was assessed using the CRU assay.^3,39  Transplantation chimeras were generated by injecting Ly5.1⁺ BM cells engineered to overexpress HOXB4(WT) or HOXB4(N→A) or control neo^r genes into lethally irradiated Ly5.2⁺ congenic recipients. At 12 weeks after transplantation, BM cells were harvested and transplanted in increasing dilutions into secondary recipients. The contribution of transplant-derived (Ly5.1⁺) transduced cells to lympho-myeloid reconstitution of secondary recipients was determined 16 weeks after transplantation by flow cytometry, and by Southern blot analyses of integrated proviruses in genomic DNA isolated from hemopoietic tissues of secondary recipients (Figure 2C-E). All integrated HOXB4(WT) (Figure 2C), HOXB4(N→A) (Figure 2D), and neo^r (Figure 2E) proviruses were free of rearrangements. There was a very good correlation observed between the percentage of donor-derived Ly5.1⁺ peripheral blood (PB) cell analysis, and the intensity of proviral signals revealed by a neo^r-specific probe in the autoradiograms (Figure 2D-E). Levels of HOXB4 expression also correlated with these parameters (not shown). These results suggested that low levels of transplant-derived reconstitution determined for HOXB4(N→A) recipients were due to the compromised function of this mutant.

The frequencies of transduced donor-derived CRUs were determined as described.^3,39  At 16 weeks after transplantation, the frequency of HOXB4(WT)⁺ CRU was 1 in 30 000 (95% confidence interval [CI]: 1 in 10 000 to 1 in 85 000), compared with 1 in 260 000 (95% CI: 1 in 93 000 to 1 in 720 000) and 1 in 350 000 (95% CI: 1 in 120 000 to 1 in 1 000 000) as determined for HOXB4(N→A)⁺ and control neo^r CRUs, respectively. Assuming a total of 2 × 10⁸ hemopoietic cells per adult mouse, the transplanted HOXB4(WT)⁺ HSCs generated a progeny of approximatively 6700 CRUs (Figure 2F). The transplanted HOXB4(N→A)⁺ and control neo^r HSCs, in contrast, generated progeny comprising approximatively 800 and 600 cells, respectively, suggesting that DNA binding is essential for HOXB4-induced in vivo expansion of HSCs, and implying a direct role for HOXB4 in transcriptional control of the gene or genes that regulate HSC self-renewal.

We previously reported that knockdown of endogenous PBX1 protein in HOXB4-overexpressing cells generated HSCs that were more than 20 times more competitive than those overexpressing HOXB4.³⁶  To assess whether this increase in repopulation potential reflected direct de-repression of HOXB4 activity, or modulation of a parallel PBX-dependent pathway or pathways involved in regulation of self-renewal, we examined the in vitro and in vivo activity of BM cells engineered to express HOXB4 carrying amino acid substitution in the interface required for interactions with PBX (YPWM→YPGM).

BM cells were infected with recombinant retroviruses encoding HOXB4(WT) and GFP, or HOXB4(W→G) and GFP, or GFP alone. In cultures initiated with sorted GFP⁺ cells, the numbers of HOXB4(W→G) and HOXB4(WT) cells increased 16- and 54-fold, respectively, above the input values compared with an approximately 10-fold increase determined for controls (Figure 3A). During a 10-day expansion period, the numbers of HOXB4(W→G) and HOXB4(WT) CFCs increased 380- and 1870-fold above the input, compared with a 7-fold increase determined for control CFCs (Figure 3B). Both HOXB4(W→G) and HOXB4(WT) also enabled the in vitro expansion of multipotent clonogenic progenitors (CFU-GEMM), which in control cultures could not be detected after day 4 (data not shown). In the in vitro competitive proliferation assays initiated with 10% GFP⁺ cells (control, HOXB4(W→G)⁺, or HOXB4(WT)⁺) and 90% nontransduced competitors, the proportions of HOXB4(W→G) and HOXB4(WT) cells increased within 11 days to 57% ± 13% and 83.5% ± 4.9, respectively, while in control cultures no increase of GFP⁺ cells above the starting 10% could be detected (Figure 3C). Together, these data indicate that HOXB4 mutant with dysfunctional PBX interacting surface retained the ability to confer the in vitro proliferation advantage to the transduced BM cells.

Three groups of BM transplantation chimeras were generated by transplanting a mixture of 10% GFP⁺ cells (HOXB4(W→G)⁺, HOXB4(WT)⁺, or control GFP⁺) and 90% nontransduced competitors in sublethally irradiated recipients. All recipients remained healthy for more than 12 months of observation. Contribution of the transplant-derived GFP⁺ cells to myeloid and lymphoid repopulation of recipients was determined 12 weeks after transplantation. In HOXB4(W→G) and HOXB4(WT) recipients (Figure 4A black and gray bars, respectively), GFP⁺ cells initially representing 10% of the graft accounted for 55% to 80% of bone marrow–derived myeloid (Mac-1⁺), 35% to 50% of splenic B-lymphoid (B-220⁺), and 60% to 80% of thymic CD4⁺CD8⁺ cell populations (Figure 4A). As expected, GFP⁺ control cells showed no proliferation advantage in this system (Figure 4A white bars). In both groups, GFP⁺ CFCs represented 45% to 55% of myeloid, and 30% to 45% of B-cell progenitor populations (Figure 4B), and no abnormalities in the relative proportions of myeloid CFCs could be detected (Figure 4C). Together, these observations suggest that HOXB4(W→G) retained the ability to confer competitive reconstitution advantage to the transduced cells without affecting their differentiation.

To verify that hemopoietic regeneration of HOXB4(W→G) and HOXB4(WT) recipients was mediated by pluripotent transduced HSCs, and to determine the numbers of cellular clones that participated in reconstitution, Southern blot analyses were performed to analyze proviral integrations in genomic DNA isolated from BM, spleen, and thymus of selected recipients (Figure 4D top panels). In individual HOXB4(W→G) and HOXB4(WT) recipients, the same proviral integration pattern could be detected in BM, spleen, and thymus, illustrating the pluripotent nature of several of the reconstituting HSC clones. In both groups, multiple proviral integrations of varying signal intensities indicated the ongoing activity of several independent clones in the indicated mice. Importantly HOXB4 overexpression was demonstrated in the hemopoietic tissues of recipients in both groups (Figure 4D bottom panel).

The numbers of transduced HSCs that reconstituted HOXB4(W→G) and HOXB4(WT) transplantation chimeras were determined using CRU assay. At 16 weeks after transplantation, the CRU frequency in recipients of HOXB4(W→G)-transduced cells was 1 in 92 000 (95% CI: 1 in 51 000 to 1 in 164 000) for a total of approximately 2200 CRU per recipient, compared with a frequency of 1 in 26 000 (95% CI: 1 in 20 000 to 1 in 36 000) and 7700 total CRUs in mice reconstituted with HOXB4(WT)-transduced cells (Figure 4E). Given that all groups of primary recipients received an estimated 15 transduced HSCs, CRU expanded by approximately 150-fold and approximately 500-fold in the HOXB4(W→G) and HOXB4(WT) chimeras, respectively. This expansion occurred in conditions that resulted in complete loss of control GFP⁺ HSCs (not shown).

Results presented in the previous section indicated that HOXB4(W→G) was less efficient in inducing HSC expansion than HOXB4(WT), suggesting a possible functional impairment of HOXB4(W→G) protein. In freshly infected cells, HOXB4(W→G) consistently appeared to be expressed at lower levels than HOXB4(WT) (Figure 5A). Importantly, HOXB4(W→G) was capable of translocating into nuclear compartments (Figure 5B), and had an intracellular half-life (70 minutes) comparable to that determined for HOXB4(WT) (Figure 5C), indicating that neither intracellular compartmentalization nor protein stability was affected by this mutation. The apparently diminished biologic activity of HOXB4(W→G) could thus simply reflect lower levels of available HOXB4(W→G) protein. To examine this possibility, we sorted HOXB4(WT)- and HOXB4(W→G)-transduced BM cells in populations expressing low or high levels of HOX proteins (GFP^lowHOXB4^low or GFP^highHOXB4^high; Figure 5D) and determined their in vitro proliferation activity. Cells expressing comparable levels of HOXB4(W→G) (Figure 5D, W→G^high, filled triangles) and HOXB4(WT) (Figure 5D, WT^low, squares) proteins were indistinguishable in their proliferation responses to cytokine stimulation as determined by ³H-thymidine incorporation (Figure 5E), and generated comparable numbers of progeny during a 7-day expansion (Figure 5F). The apparently diminished potential of HOXB4(W→G) thus likely reflected lower levels of available HOXB4(W→G) protein, and not functional impairment resulting from the loss of ability to interact with PBX in DNA binding. To complement these experiments and further validate this point, in vitro competitive assays were initiated with mixturecomprising20%ofHOXB4^HighGFP^HighLy5.1⁺and80%HOXB4^Low-GFP^LowLy5.2⁺. The proportion of Ly5.1⁺ (HOXB4^high) cells increased over time in culture, and represented by day 18 approximately 60% of the total cell population (Figure 5G), indicating that an increase in HOXB4 levels enhanced the competitive proliferation advantage of the transduced cells.

Together, the results presented in this report show that the capacity of HOXB4 to enhance HSC proliferation depends on its ability to bind DNA. Our observations also indicate that HOXB4-induced HSC expansion represents a PBX-independent function of this transcription factor, and establish a strong correlation between protein levels of HOXB4 and in vitro biologic activity. These studies also validate future efforts to identify HOXB4-dependent target genes involved in HSC activity.

Results presented in this paper indicate that the DNA-binding ability of HOXB4 is essential for its HSC-expanding function and suggest that HOXB4 acts as a transcriptional regulator of genes that govern self-renewal. These observations are consistent with results of studies showing that the transforming potentials of NUP98-Hoxa9⁴⁶  and NUP98-Hoxd13⁴⁷  are absolutely dependent on the integrity of their respective homeodomains. Moreover, these homeodomains fused to NUP98 retained their leukogenic potential,⁴⁸  indicating that upon acquisition of transactivating function provided by NUP98, homeodomains were sufficient for activation of Hox-responsive genes.

In transactivation assays, only a weak suppressive effect of HOXB4 on expression of reporter genes driven by synthetic promoters comprising HOX target sites could be detected (M. Featherstone, unpublished observations, June 2001). In agreement with the repressive role proposed for HOX4 paralogs,¹⁷  HOXB4 suppressed transcription of Ras antagonist X-Rap1 during Xenopus embryonic development,⁴⁹  and repressed expression of c-MYC in hemopoietic cell line HL-60.^50,51  However, HOXB4 also acted as a positive regulator of Drosophila Iroquois 5⁵²  and Flash transcription,⁵³  indicating that cellular context plays a major role in defining the nature of HOXB4 regulatory activity.

HOXB4 could thus promote HSC self-renewal by suppressing transcription of genes implicated in commitment and differentiation, or by enhancing expression of a distinct element or elements that enables self-renewal divisions and maintenance of “stemness.” To discriminate between these 2 possibilities we generated chimeras comprising HOXB4 and engrailed repression domain,⁵⁴  or HOXB4 and VP16 activation domain,^54,55  and examined their activity in the mouse BM model system (N.B. and M.L., unpublished observations, February 2004). Preliminary results suggest that addition of strong repressor domain abolished the ability of HOXB4 to confer the in vitro and in vivo competitive proliferation advantage to the transduced BM cells. In contrast, addition of VP-16 transactivating domain failed to enhance activity of HOXB4 above the levels determined for HOXB4(WT), implying that HOXB4 acts as a positive regulator of a stem cell–specific gene or genes. Moreover, in the transduced BM populations VP-16/HOXB4 induced a major up-regulation of endogenous HOXB4 protein, suggesting an evolutionally conserved HOXB4 auto-regulatory loop.⁵⁶  An increase in HOXB4 levels has also been reported in conjunction with Tpo-promoted⁵⁷  and Wnt3a-promoted⁵⁸  HSC self-renewal, implying that up-regulation of HOXB4 could represent an initiating event in this process.

We reported previously that HOXB4 and PBX cooperated in enhancing proliferation in a fibroblast model system.³³  PBX knockdown, in contrast, significantly enhanced the competitive repopulation capacity of HOXB4-transduced HSCs,³⁶  suggesting that requirement for HOX-PBX interactions depends on the cellular context. The latter study relied on examination of genetic interactions between HOXB4 and PBX1, and could not exclude the possibility that other members of the TALE family compensated for the absence of PBX1 in HOXB4-promoted HSC expansion. However, the HOXB4(W→G) mutant examined in this study retained the HOXB4-specific stem cell–expanding potential despite the W→G substitution that prevented the cooperative HOX/PBX DNA binding. The Hoxa9-induced HSC expansion during the preleukemic phase³⁸  is comparable to expansions determined for HOXB4. Interestingly, the Hoxa9 mutant lacking the ability to interact with PBX was capable of immortalizing myeloid CFCs,²⁸  and NUP98-HOXA9 carrying a W→G substitution enhanced proliferation of the transduced BM cells³⁴  and induced leukemic transformation (E. Kroon and G.S., manuscript in preparation). Direct HOX/PBX interactions thus appear to be dispensable in the context of HOX-promoted stem cell proliferation.

Differences in protein expression levels in cells transduced with wild-type versus W→G mutant constructs remain uncertain. Our observations indicate that there was no difference in titers between the 2 retroviruses and the half-life of both wild-type and mutant W→G proteins are comparable (Figure 3C). This, combined with the results indicating that mRNA levels are comparable between bone marrow cells transduced with wild-type versus mutant Hoxb4 cDNA, suggest that the W→G point mutation may interfere with Hoxb4 translation.

Based on our recent studies demonstrating that down-regulation of PBX1 significantly enhanced the competitive activity of HOXB4-transduced BM cells,³⁶  the expectation was that our HOXB4(W→G) mutant would behave as a “super HOXB4” protein. Although we cannot formally exclude the possibility of some in vivo HOXB4-PBX interactions involving nonidentified HOX domains as suggested with HOXA9,²⁸  it appears that the likely explanation for these observations is that HOXB4 and PBX genes act on distinct pathways in these cells.

In summary, studies presented in this report show that the homeodomain-mediated HOXB4-DNA interactions are essential for a HOXB4-induced increase in the self-renewal activity of HSCs and indicate that direct HOXB4-PBX interactions involving the YPWM motif are not required for this function. In future studies, these observations will form the basis for the identification of HOX-activated genes contributing to HSC self-renewal.

We thank Mrs Melanie Frechette for her expertise and help regarding maintenance of animals. The help of Mr Eric Massicotte and Mrs Martine Dupuis with flow cytometry is also acknowledged.