Differential and Common Leukemogenic Potentials of Multiple NUP98-Hox Fusion Proteins Alone or with Meis1

Abstract

NUP98-Hox fusion genes are newly identified oncogenes isolated in myeloid leukemias. Intriguingly, only Abd-B Hox genes have been reported as fusion partners, indicating that they may have unique overlapping leukemogenic properties. To address this hypothesis, we engineered novel NUP98 fusions with Hox genes not previously identified as fusion partners: the Abd-B-like gene HOXA10 and two Antennepedia-like genes, HOXB3 and HOXB4. Notably, NUP98-HOXA10 and NUP98-HOXB3 but not NUP98-HOXB4 induced leukemia in a murine transplant model, which is consistent with the reported leukemogenic potential ability of HOXA10 and HOXB3 but not HOXB4. Thus, the ability of Hox genes to induce leukemia as NUP98 fusion partners, although apparently redundant for Abd-B-like activity, is not restricted to this group, but rather is determined by the intrinsic leukemogenic potential of the Hox partner. We also show that the potent leukemogenic activity of Abd-B-like Hox genes is correlated with their strong ability to block hematopoietic differentiation. Conversely, coexpression of the Hox cofactor Meis1 alleviated the requirement of a strong intrinsic Hox-transforming potential to induce leukemia. Our results support a model in which many if not all Hox genes can be leukemogenic and point to striking functional overlap not previously appreciated, presumably reflecting common regulated pathways.

The clustered homeobox Hox genes encode a highly conserved family of transcription factors characterized by a 60-amino-acid DNA binding motif, the homeodomain. While Hox genes were first recognized for their prominent roles in embryonic development, their involvement in both normal and malignant hematopoietic processes is now well documented (2, 6, 30).

A central role for Hox genes in hematological malignancies is supported by the frequently observed elevation of Hox gene expression in acute myeloid leukemia (AML) (11, 14, 20) and the common involvement of their upstream regulator, MLL, in both myeloid and lymphoid leukemia (8). A more direct involvement for Hox genes in leukemia is supported by their frequent fusion to the nucleoporin gene NUP98. Hox genes are the most frequent fusion partner in an ever-growing number of NUP98 fusion genes isolated almost exclusively in patients with myeloid leukemia (AML, posttherapy AML, and chronic myeloid leukemia). Reported Hox partners of NUP98 include HOXA9 (3, 21), HOXD13 (27), HOXA13 (9), HOXA11 (9), HOXC11 (35), and HOXC13 (19). All NUP98-Hox fusions reported to date consist of the N terminus of NUP98, containing a region of multiple phenylalanine-glycine (FG) repeats that may act as a transcriptional coactivator through binding to CBP/p300 (13), and the C terminus of the Hox gene product, containing the intact homeodomain and various extents of flanking amino acids. The leukemogenic potential of such chimeric proteins was confirmed for NUP98-HOXA9 (NA9) and NUP98-HOXD13 (ND13) in retrovirally transduced murine bone marrow transplantation models with both genes inducing AML and/or a myeloproliferative syndrome (16, 25). The transforming potentials of NA9 and ND13 were found to be dependent on the ability of their homeodomains to bind DNA (13, 25), and the Pbx interacting motif of HOXA9 in NA9 was found to be dispensable, suggesting that the Hox DNA-binding domain is the minimal and possibly only significant contribution of the Hox fusion partner.

The leukemogenic activity of Hox genes is not restricted to chimeric NUP98-Hox proteins, since HOXB8, HOXA10, HOXB3, and HOXA9 have previously been shown to induce AML upon engineered overexpression (15, 24, 31, 37). This suggests that a key result of NUP98 fusion with Hox is deregulated expression of Hox-responsive genes by the chimeric Hox protein, in which NUP98 replaces the transcriptional activity of the Hox N-terminal region (13). Consistent with this, both HOXA9 and NA9 studied in similar experimental settings are leukemogenic and block myeloid progenitor differentiation in interleukin-3 and granulocyte-macrophage colony-stimulating factor (7, 16). Nevertheless, differences in disease latency, cooperativity with Meis1, and other in vitro growth properties suggest that NUP98-Hox fusions have different properties compared to their normal counterparts. For instance, in vitro, HOXA9 did not replicate the effect induced by NA9, such as proliferation in stem cell factor and granulocytic differentiation block with granulocyte colony-stimulating factor (7).

Intriguingly, so far only Hox genes belonging to the Abd-B-like Hox paralog groups (groups 9 to 13) have been characterized as NUP98 fusion partners, and within those, only members from paralog groups 9, 11, and 13 have been observed. Thus, it may be that only a subset of Hox genes, sharing potential overlap in gene targets and/or pathways, can complement NUP98. Outside of the Abd-B-like Hox genes, it is of interest that HOXB4, while having potent ability to enhance hematopoietic stem cell (HSC) expansion is on its own nonleukemogenic, whereas the near neighbor HOXB3 can cause marked myeloproliferation and ultimately leukemia (1, 31). These findings raise the interesting question as to the intrinsic leukemogenic potential of Hox genes across the clusters and to their leukemogenic potential in the context of fusion genes.

The leukemic potential of intact Hox genes, such as HOXB3 and HOXA9, and NUP98 fusions with HOXA9 or HOXD13 has been shown to be increased by co-overexpression of Meis1 (15, 22, 25, 36) as evidenced by decreased disease latency. The basis for this cooperativity is unclear because, although in vitro studies have suggested that Meis1 expression can synergize with Hox genes in promoting a stronger differentiation block of Hox-transformed myeloid progenitors (5), it does not show leukemogenic activity on its own in vivo (16, 25). Understanding the role of Meis1 is further complicated by the fact that several of the known collaborating Hox proteins involved lack the ability to physically interact with Meis1, as in the case of HOXB3 and ND13. Finally, it is not clear whether all NUP98-Hox fusions could be strongly complemented by Meis1 since the leukemogenic potential of NA9 was only weakly augmented by Meis1, whereas Meis1 strongly complemented ND13 (16, 25).

The principal objectives of the work described here were to test whether the preferential identification of Abd-B-like Hox genes as NUP98 fusion partners in human leukemias is a consequence of their special intrinsic properties or, on the contrary, whether other Hox genes might share the functional capacity to form oncogenic NUP98 fusion genes. The leukemogenic potential of various NUP98-Hox fusion genes representing Hox genes of different paralog groups and different leukemogenic potentials was tested in vivo, alone and in concert with Meis1, to enlarge the knowledge of the role of Meis1 in Hox-related leukemia. Furthermore, in order to gain further insight into the mechanism(s) of Hox-mediated leukemia, we analyzed the impact of the native and NUP98-fused Hox genes on bone marrow differentiation and proliferation.

MATERIALS AND METHODS

Retroviral vectors and engineering of novel NUP98-Hox fusion genes.

The murine stem cell virus (MSCV) Flag-NUP98-HOXD13 (ND13) internal ribosomal entry site (IRES)-enhanced green fluorescent protein (eGFP), MSCV ND13 IRES-eGFP, MSCV Meis1 IRES-yellow fluorescent protein (YFP), MSCV HOXB4 IRES-eGFP, and MSCV HOXA10 IRES-eGFP viral vectors used were previously described (1, 4, 25). We engineered the fusion of the cDNA portion corresponding to the homeobox-containing exon of HOXA10 (NUP98-HOXA10; NA10), HOXB4 (NB4), and HOXB3 (NB3) with the NUP98 portion found in the ND13 fusion gene, which consists of exons 1 to 12 of NUP98 (Fig. 1A). All gene fragments were obtained by PCR from cloned cDNAs with the Platinum Taq DNA polymerase High Fidelity (Invitrogen, Burlington, Canada); the primer sequences are available in the appendix. Constructs were validated by sequencing and correct expression and transmission were confirmed by Western blot and Southern blot analysis (Fig. 1B and C).

FIG. 1.

Engineering of novel NUP98-Hox fusion genes. (A) Three new NUP98-Hox fusion genes were engineered by fusing the cDNA sequence corresponding to the homeobox-containing exon of HOXA10 (NA10), HOXB4 (NB4), and HOXB3 (NB3) to that of NUP98. Only NA10 retained its Pbx-interacting motif, which is indicated as a black rectangle. The left grey boxes indicate the Flag tag, and the homeodomains are illustrated as light grey rectangles. (B) Western blot analysis of cell extracts from transduced BaF3 cells with a monoclonal Flag antibody. (C) Southern blot analysis showing full-length proviral integration of the various NUP98-Hox IRES-GFP viral vectors in transduced bone marrow cells. (D) NA10 expression provides bone marrow cells with a growth advantage in vitro. The proportion of transduced GFP⁺ bone marrow cells over time in liquid culture established with nonpurified transduced bone marrow cells is shown. Results of a representative experiment are shown: mean ± standard deviation of triplicate culture (n = 3). (E) Expression level of the GFP reporter gene in circulating leukocytes from mice transplanted with NB4, NA10, NB3, and ND13 transduced bone marrow cells 8 to 13 weeks posttransplant. Histogram profiles of representative and GFP control (CTL) mice are shown.

Mice and retroviral infection of primary bone marrow cells.

Parental strain mice were bred and maintained at the British Columbia Cancer Research Centre animal facility. Donors of primary bone marrow cells were (C57BL/6Ly-Pep3b × C3H/HeJ)F₁ mice and recipients were (C57BL/6J × C3H/HeJ)F₁ (B6C3) mice. Primary mouse bone marrow cells were transduced as previously described (25). Briefly, bone marrow cells were harvested from mice treated 4 days previously with 150 mg of 5-fluorouracil/kg (Faulding, Underdaler, Australia) and stimulated for 48 h in Dulbecco's modified Eagle's medium supplemented with 15% fetal bovine serum, 10 ng of human interleukin-6 per ml, 6 ng of murine interleukin-3 per ml, and 100 ng of murine stem cell factor (StemCell Technologies Inc., Vancouver, Canada) per ml. All bone marrow cultures used medium. The cells were infected by cocultivation with irradiated (4,000 cGy of X-ray) GP+ E-86 viral producer cells with the addition of 5 μg of protamine sulfate (Sigma, Oakville, Canada) per ml.

A 1:1 mixture of producers of the NUP98-Hox viruses (with the green fluorescent protein [GFP] marker) and Meis1 virus (with the YFP marker) was used to coinfect bone marrow cells. Loosely adherent and nonadherent cells were harvested from the cocultures after 2 days and cultured for 48 h in the same medium without protamine sulfate. Where indicated, transduced cells were highly purified based on expression of GFP, YFP, or both fluorescent proteins with a FACS-Vantage (Becton Dickinson, Mississauga, Canada), as previously described (25).

Bone marrow transplantation and monitoring of recipients.

Purified GFP⁺ and/or nonpurified (unsorted cells) transduced bone marrow cells were injected into the tail vein of irradiated (900 cGy of ¹³⁷Cs γ-radiation) recipient F₁ B6C3 mice. Mice transplanted with singly transduced bone marrow cells received from 1.4 × 10⁵ to 3.5 × 10⁵ GFP⁺ cells/transplant and mice injected with doubly transduced cells received from 0.2 × 10⁴ to 8.0 × 10⁴ GFP⁺/YFP⁺ bone marrow cells. For peripheral blood, bone marrow, and spleen cell analysis, single-cell suspensions were stained with the following monoclonal antibodies: phycoerythrin-labeled Gr-1, Mac-1, B220, Ter-119, and c-Kit (all obtained from Pharmingen, San Diego, Calif.). The immunoglobin E (IgE) receptor was detected with a polyclonal biotinylated anti-IgE receptor antibody (StemCell Technologies Inc.) followed by streptavidin-phycoerythrin (Pharmingen, San Diego, Calif.). Morphological analysis of peripheral blood, bone marrow, and spleen cells and histological analysis were performed as previously described (25).

In vitro assays.

Cell proliferation was assessed in the same media described above. Differentiation of clonogenic progenitors was analyzed by plating cells in 1 ml of methylcellulose culture medium per petri dish in standard conditions (Methocult M3434, StemCell Technologies Inc.), containing 10 ng of murine interleukin-3 per ml, 10 ng of human interleukin-6 per ml, 50 ng of murine stem cell factor per ml, and 3 U of human erythropoietin per ml. Colonies were scored microscopically with standard criteria after 8 to 10 days.

CFU-S assay.

GFP⁺ purified bone marrow cells were first cultured for 7 days before injection. The total cell progeny from the starting equivalent (day 0) of 6 cells up to 1 × 10⁶ cells were then injected into irradiated recipient mice. The recovery of CFU-spleen (CFU-S) cells was measured by determining the number of macroscopic colonies on the spleen at day 12 postinjection after fixation in Telleyesnickzky's solution.

Southern blot analysis.

Genomic DNA was isolated with DNAzol reagent as recommended by the manufacturer (Invitrogen), and Southern blot analysis was performed as previously described (29). The stable integration of the provirus was confirmed by digesting the DNA with NheI (cutting in each long terminal repeat) and then the DNA-containing membrane was probed with ³²P-labeled GFP cDNA. For clonality analysis DNA from NUP98-Hox plus Meis1 primary or secondary transplant recipients was digested with EcoRI and probed with a ³²P-labeled NUP98 cDNA. DNA from HOXB4-plus-Meis1- and HOXA10-plus- Meis1-transplanted mice was digested with XhoI and EcoRI and hybridized to HOXB4- or HOXA10-specific probes, respectively.

Western blot analysis.

Transduced BaF3 cells (infected with the same procedure as that for bone marrow) were lysed (400 mM NaCl, 0.1% Triton, 0.1 M HEPES, pH 7.5). Cell lysate was loaded onto a 10% Bis-Tris polyacrylamide gel (NuPAGE [NP0303], Invitrogen) and blotted to a Biotrace polyvinylidene difluoride membrane (Pall Corporation, Ann Arbor, Mich.). A monoclonal anti-Flag (M2, Sigma) and a donkey horseradish peroxidase-conjugated anti-mouse immunoglobulin antibody were then used (Jackson ImmunoResearch Lab. Inc., West Grove, Pa.). Protein expression was detected with an enhanced Luminol reagent (Renaissance, Boston, Mass.).

Statistical analysis.

Data were statistically tested with the t test for dependent or independent samples (Microsoft Excel). Differences with P values <0.05 were considered statistically significant.

RESULTS

Redundancy of the ability of Abd-B-like Hox genes to cause AML as NUP98 fusion partners.

The increasing list of Abd-B-like Hox genes that have been identified fused to NUP98 in human AML strongly suggests a functionally equivalent role for this closely related group of genes in leukemia. Interestingly, however, this list does not include any of the Hox genes belonging to the 10th paralog group, even though HOXA10 has been reported to be leukemogenic in mouse models. This raises the possibility that Hox genes of the 10th paralog group do not possess the properties required to form leukemogenic NUP98 fusion genes. To clarify this issue, we engineered a novel NUP98-Abd-B fusion gene, NUP98-HOXA10 (NA10), and studied the effect of its overexpression in a mouse bone marrow transplantation model. To construct NA10, the cDNA portion corresponding to the second exon of HOXA10, containing its homeodomain and Pbx-interacting motif, was fused in frame to the NUP98 portion found in ND13, thus recapitulating a typical NUP98-Hox fusion gene (Fig. 1). NA10 cDNA was cloned into the MSCV-IRES-GFP retroviral vector, in which an internal ribosomal entry site allows the cotranslation of the GFP reporter gene and NA10.

Expression of NA10 provided freshly transduced bone marrow cells with a strong growth advantage, as evidenced by the increasing proportion of transduced GFP⁺ cells over time in liquid culture (Fig. 1D). This effect was similar to that induced by ND13 and provided initial support for HOXA10 as an oncogenic NUP98 fusion partner. To determine whether the growth advantage could translate into leukemogenic activity in vivo, lethally irradiated mice were reconstituted with NA10-transduced bone marrow cells.

NA10-transduced bone marrow cells reconstituted the hematopoietic system of irradiated mice efficiently, with the proportion (Table 1) and expression levels of GFP⁺ mononuclear cells (Fig. 1E) similar to that observed for ND13 control-transplanted mice. No significant differences were observed in the number of white blood cells and red blood cells in the NA10 and control HOXA10 and GFP recipients 8 weeks posttransplant (Table 1). However, as first detected at 24 weeks, three out of eight healthy-looking NA10 mice displayed elevated white blood cell numbers (14 × 10⁶ to 35 × 10⁶/ml) due to increased myeloid cells (Gr1⁺ and/or Mac1⁺), consistent with a mild myeloproliferative condition, as previously observed in some ND13 and NUP98-HOXA9 mice (16, 25).

TABLE 1.

Engraftment levels in mice transplanted with singly transduced bone marrow cells 8 to 13 week posttransplant^a

Gene transduced	% GFP+ white blood cells ± SD	No. of blood cells
		White (106/ml) ± SD	Red (109/ml) ± SD
GFP (control)	63.5 ± 11.5 (8)b	7.9 ± 1.5 (6)	10.8 ± 1.0 (7)
NA10	70.7 ± 5.9 (7)	7.7 ± 2.0 (7)	11.3 ± 1.6 (3)
NB4	69.5 ± 5.6 (10)	7.8 ± 1.3 (10)	9.8 ± 1.0 (9)
A10	50.7 ± 14.5 (7)	9.0 ± 3.6 (7)	10.7 ± 1.1 (3)
B4	63.6 ± 10.3 (6)	6.9 ± 1.3 (6)	9.1 ± 1.7 (6)

^aResults shown are means from at least two independent cohorts.

^bTotal number of recipients analyzed.

Recipients transplanted with NA10-transduced bone marrow cells had a much reduced survival rate compared to GFP controls but similar to that observed for control ND13 mice (Fig. 2A), with a median survival of 223 days. Examination of sick NA10 recipients revealed that they had elevated white blood cell counts (3- to 50-fold above normal), anemia, and splenomegaly (Table 2). An AML phenotype was confirmed by Wright-Giemsa staining of bone marrow cytospins, which revealed a high proportion (≥25%) of blast cells in the marrow of the sick NA10 recipients (Fig. 3A; Table 2). The myeloid nature of the leukemic cells was confirmed by flow cytometry (Fig. 2B), which revealed that the majority of the leukemic bone marrow cells stained positive for Mac1, Gr1, and c-Kit, although a small degree of phenotypic heterogeneity was observable among different mice. This phenotypic variability likely reflects small differences in the predominant stage of differentiation impairment influenced by the level of NA10 expression or the nature of an extra mutation(s) acquired during leukemia progression (see below).

FIG. 2.

Novel fusion gene NUP98-HOXA10 but not NUP98-HOXB4 induces leukemia in mice. (A) Survival curve of mice transplanted with NA10 (n = 8), NB4 (n = 16), ND13 (n = 13), and GFP (n = 12) transduced bone marrow. (B) Immunophenotyping of leukemic cells. Flow cytometry profiles of bone marrow cells for three sick NA10 recipients and one control mouse (CTL) are shown. Expression of GFP is shown on the x axis and that of the cell surface antigens on the y axis. Uns, unstained.

TABLE 2.

Hematological characteristics of diseased NA10 mice

Mice (no.)	Mean day of sacrifice ± SD	Mean no. of white blood cells (106/ml) ± SD	Mean no. of red blood cells (109/ml) ± SD	Mean spleen wt (g) ± SD	Mean % blasts in marrow ± SD
NA10 (5)	257 ± 70.2	113.5 ± 149	2.7 ± 1.8	0.59 ± 0.27	36 ± 9
Normal B6C3 control (3)		7.1 ± 6.6	6.8 ± 1.4	0.10 ± 0.02	8.2 ± 2.7

FIG. 3.

Wright-Giemsa staining of leukemic bone marrow cells harvested from moribund NUP98-HOXA10, NUP98-Hox/Meis1, and HOX/Meis1 mice. Staining of leukemic bone marrow cells harvested from moribund NA10 mice (A) or from moribund NUP98-Hox/Meis1 and Hox/Meis1 mice (B). M, Meis1.

All leukemias originated from transduced bone marrow cells, as confirmed by GFP expression (Fig. 2B) and Southern blot analysis, which detected the presence of the intact NA10 provirus (data not shown). Histopathological analysis confirmed significant infiltration of immature myeloid cells in the spleen, liver, kidneys, and occasionally in lungs (data not shown). The leukemia was transplantable, and all secondary recipients injected with as few as 5,000 bone marrow cells also developed AML, with a latency ranging from 31 to 69 days (three primary mice tested). The much-reduced disease latency seen in the secondary transplants argues that additional mutations had been acquired during the genesis of the leukemias. As previously reported and confirmed here, control HOXA10 (data not shown) and ND13 recipients also developed AML over time (25, 37). In contrast to the reported disease kinetics observed between HOXA9 and NA9 mice (16), HOXA10, associated with a median survival of 151 days, induced AML faster than NA10. Together, these results demonstrate that the Abd-B-like HOXA10 gene is a potent oncogenic NUP98 fusion partner and support a model in which all members of this class of Hox genes have overlapping leukemogenic properties.

Fusion of the nonleukemogenic HOXB4 gene to NUP98 does not cause AML.

The results presented above supported the hypothesis that Abd-B-like Hox genes have common transforming properties that can be complemented by fusion to NUP98. However, it remained possible that any Hox gene, regardless of its intrinsic leukemogenic potential, could form an oncogenic fusion gene with NUP98. To investigate this, we engineered a fusion between NUP98 and HOXB4, a Hox gene previously found to be nonleukemogenic in transplant models and not yet implicated in human leukemia.

In contrast to NA10, NB4 did not induce leukemia even after follow-up of 1 year (Fig. 2A). This lack of leukemogenic activity was not attributable to low-level engraftment or expression levels, as evident from GFP expression in peripheral blood mononuclear cells (Table 1 and Fig. 1E). During that observation period, three NB4 and two GFP recipients died of unknown causes with no evidence of myeloproliferative disorder or hematological abnormalities. Clonogenic progenitor assays and fluorescence-activated cell sorter (FACS) analysis of the marrow of one of the NB4 mice revealed no apparent perturbation, and all secondary bone marrow transplants remained healthy (>47 weeks). These results indicate that the fusion of NUP98 to HOXB4 was without detectable leukemogenic activity and thus suggest that not all Hox genes can form leukemogenic NUP98 fusions.

To discriminate between the possibilities that HOXB4 could not complement NUP98 due to lack of intrinsic leukemogenic activity or that only Abd-B genes are able to collaborate with NUP98, we engineered a fusion between NUP98 and another Antennapedia-like Hox gene, HOXB3, that on its own has been shown to induce AML. Low viral titers resulted in sufficient numbers of GFP⁺ NB3 cells to transplant only two mice; nevertheless, one of two NB3 recipients developed leukemia, as evidenced by a high white blood cell count (110 × 10⁶/ml), low red blood cell count (5.3 × 10⁹ red blood cells/ml), splenomegaly (0.58 g), and a proportion of blasts of 21% in its marrow at ≈235 days posttransplant, a latency similar to that reported for the intact HOXB3 gene (31). Together, these data suggest that the ability of Hox genes to form leukemogenic fusions with NUP98 is not restricted to Abd-B-like Hox genes, but may rather be determined by the intrinsic leukemogenic potential of the Hox gene fused.

Redundant ability of Hox and NUP98-Hox genes to cause AML in collaboration with Meis1 is independent of their intrinsic leukemogenic potential.

The Hox cofactor Meis1 has been shown to reduce the time of disease onset for several Hox-induced leukemias, suggesting redundancy in the ability of Hox genes to collaborate with Meis1. A second important observation arising from these studies is that the strength of this cooperation seems to correlate with the intensity of the leukemogenic potential of the Hox gene involved. In order to clarify these issues, we investigated the potential synergy between Meis1 and NA10 (a strong oncogene), NB3, and NB4 (nononcogenic). In addition, to determine whether the synergy between Meis1 and the fusion genes reflects an intrinsic property inherited from their normal Hox counterparts, both HOXA10 and HOXB4 were coexpressed with Meis1 in vivo.

Bone marrow cells were cotransduced with the various NUP98-Hox-IRES-GFP and Meis1-IRES-YFP retroviruses and injected into irradiated mice. Strikingly, all fusion and Hox genes tested collaborated with Meis1 to induce leukemia, as evidenced by reduced disease latency or, in the case of NB4 and HOXB4, conversion to leukemogenic activity (Fig. 4A and B). Expression of Meis1 alone did not cause disease, as previously reported (15). Moreover, the median survival seen with Meis1 in combination with NA10 or HOXA10 was much shorter than in combination with NB3, HOXB4, or NB4 (Fig. 4B), reflecting the previously documented different leukemogenic potencies of the Hox genes by themselves. Moreover, although as single factors NB4 and HOXB4 were unable to induce leukemia, both genes were equally capable of inducing the disease with indistinguishable latency in combination with Meis1 (Fig. 4A and B; Table 3). This ability of nonleukemogenic Hox genes to cause AML in collaboration with Meis1 further supports the redundancy of such collaboration.

FIG. 4.

Coexpression of Meis1 strongly increases the leukemogenic potential of NA10 and induces AML with NB4, NB3, and two native Hox genes. (A) Survival curve of mice transplanted with unsorted or enriched cotransduced bone marrow cells (see Materials and Methods for details). (B) The median disease latency for HOXA10 plus Meis1 (M) and NA10 plus Meis1 is shorter than that for HOXB4 plus Meis1, NB4 plus Meis1, or NB3 plus Meis1. The t test P values are also shown.

TABLE 3.

Hematological characteristics of moribund mice transplanted with bone marrow cells cotransduced with Meis1 and Hox or NUP98-Hox genes

Genes transduced (no. of mice)	Mean time of sacrifice (days posttransplant) ± SD	Mean no. of blood cells ± SD		Mean spleen wt (g) ± SD	% Bone marrow cells ± SD
		White (106/ml)	Red (109/ml)		Blasts	Myeloblasts
A10/M (3)	106.7 ± 27.7	115.8 ± 89.4	1.95 ± 1.0	0.56 ± 0.16	23.3 ± 5.7	55.7 ± 2.5
NA10/M (3)	69.7 ± 7.5	140 ± 133	5.5 ± 0.9	0.41 ± 0.11	32.0 ± 8.9	50.7 ± 4.7
B4/M (4)	144.3 ± 9.2	131.3 ± 63.7	4.0 ± 2.4	1.0 ± 0.5	28.0 ± 9.1	53.8 ± 3.8
NB4/M (4)	167.8 ± 21.0	65.5 ± 45.0	5.1 ± 2.9	0.7 ± 0.4	20.0 ± 6.0	46.0 ± 8.5
Normal B6C3 (3)		7.1 ± 6.6	6.8 ± 1.4	0.10 ± 0.02	8.2 ± 2.7	10.4 ± 4.1

The phenotype of the diseased NA10/Meis1 (NA10/M), A10/M, NB4/M, and B4/M mice closely resembled that induced by NA10 and that previously reported from ND13/M (25), with elevated white blood cell counts, anemia, splenomegaly, and a high percentage of poorly differentiated myeloid blasts in the bone marrow (Fig. 3B; Table 3). The myeloid phenotype of the leukemic cells was confirmed by flow cytometry (predominantly Gr-1⁺ Mac-1⁺ B220⁻ CD4/8⁻, data not shown). Importantly, flow cytometric data from representative mice (Fig. 5A) revealed that leukemic bone marrow cells were positive for coexpression of the transduced genes, as evidenced by the high proportion of GFP/YFP double-positive cells. The origin of the leukemias from cotransduced cells, initially a minority population (as low as 0.4%) in the transplant inocula, was further confirmed by Southern blot analysis, showing the presence of NUP98-Hox and Meis1 or Hox and Meis1 proviruses in primary and secondary leukemic mice (Fig. 5B).

FIG. 5.

NUP98-Hox/Meis1- and Hox/Meis1-induced myeloid leukemias are of clonal or oligoclonal origin. (A) Analysis of GFP and YFP expression in leukemic bone marrow cells (representative profiles shown). Fresh normal marrow (control, CTL) and GFP- and YFP-transduced control bone marrow cells are also shown. (B) Southern blot analysis from genomic DNA harvested from the leukemic marrow confirmed the presence of the Meis1-IRES-YFP and NUP98-Hox-IRES-GFP provirus in the diseased mice. (C) To estimate the frequency of leukemia-initiating cells and analyze the clonal origin of the leukemias, bone marrow from moribund primary recipients was injected into irradiated secondary recipients at various doses. The lanes loaded with the primary recipient genomic DNA are indicated by P, and the lanes indicated by the numbers 1, 2, 3, and 4 correspond to DNAs of the secondary transplant recipients which received 5 × 10⁵, 5 × 10⁴, 5 × 10³, and 500 bone marrow cells, respectively, from the primary recipients. The arrows indicate the locations of endogenous Hox or Meis1 sequences hybridizing to the probes.

The leukemias caused by NA10/M, A10/M, NB4/M, and B4/M were transplantable, with secondary transplant recipients succumbing to leukemia in 20 to 70 days posttransplant. The frequency of the leukemic initiating cell was estimated by injecting different doses of primary leukemic marrow cells into multiple recipients. All mice from all gene combinations died after receiving as few as 5,000 cells, and the majority of mice receiving 500 cells also succumbed to AML (Table 4), indicating a high frequency of leukemia-initiating cells. Clonal analysis of genomic DNA harvested from the primary and secondary transplants revealed that the majority of the leukemias were of clonal and/or oligoclonal origin, as evidenced by the same banding pattern of equal intensity in the primary and secondary transplants (Fig. 5C). Interestingly, some sick NB4/M and B4/M mice showed a polyclonal bone marrow reconstitution at the time of analysis, as evidenced by the presence of multiple NB4 or B4 proviral bands in the primary recipients (see NB4/M 10.1 and 15.1 and B4/M 12.1 lanes in Fig. 5C), which is consistent with HOXB4's intrinsic property of enhancing HSC regeneration capacity, a property that seems preserved in NB4.

TABLE 4.

Leukemias induced by NUP98-HOX fusion genes plus Meis1 are highly transplantable

Gene combination (no. of mice tested)	% of mice surviving 125 days after transplantation with indicated no. of leukemic bone marrow cells:		Median survival (days) of mice receiving 5,000 leukemic cells
	500	5,000
NA10/Meis1 (4)	33	0	19.5
A10/Meis1 (4)	25	0	45
ND13/Meis1 (3)	0	0	27
NB4/Meis1 (4)	25	0	46
B4/Meis1 (4)	50	0	69

Previous studies of ND13 revealed that its growth-promoting activity was dependent on an intact homeodomain. These results suggested that the homeodomain might possibly be the most significant contribution of the Hox partner. To determine whether Meis1 can collaborate with NUP98 fusion genes containing only the homeodomain encoded by the Hox gene, a series of mice were transplanted with bone marrow cells cotransduced with Meis1 and ND13hd or NA10hd, in which only the homeodomain (hd) encoded by HOXD13 (amino acids 265 to 325) or HOXA10 (amino acids 324 to 385) was fused in frame to NUP98. As seen in Fig. 4A, ND13hd was as efficient as ND13 in inducing leukemia in collaboration with Meis1, and though following a longer latency, all mice transplanted with NA10hd/Meis1-transduced cells also succumbed to leukemia, indicating that the Hox homeodomain alone, for these genes, is enough for collaboration with Meis1.

Strong leukemogenic potential of Abd-B-like Hox genes correlates with their robust ability to promote cell growth and block differentiation.

Based on the results described above, we speculated that the high leukemogenic potential of Abd-B-like Hox genes might be explained by their strong ability to promote cell proliferation and/or block differentiation.

All three novel fusion genes enhanced the proliferation of transduced bone marrow cells in short-term liquid culture, as evidenced by the increased proportion of GFP⁺ cells over time (Fig. 1D and data not shown). The total expansion of GFP⁺ cells (i.e., total number of GFP⁺ cells after 14 days/input GFP⁺ cells) induced by NA10 was, however, consistently greater (10- to 50-fold higher) than that induced by NB3 and NB4 and similar to that obtained for ND13 (Fig. 6A). Moreover, this differential growth-promoting effect induced by the various fusion genes was also seen in more primitive cells, as shown by day 12 CFU-S. As previously reported, GFP cultures yielded a limited number of day 12 CFU-S cells following 1 week of culture. In contrast, all fusion genes tested produced a prominent increase, an average of 4,700-fold, in the yield of CFU-S-like cells (Fig. 6B). Again, the magnitude of the impact was greater for NA10, which produced the highest day 12 CFU-S recovery, matching that observed with ND13 (Fig. 6B), whereas NB4 consistently produced the smallest yield in all experiments (n = 3), followed by NB3.

FIG. 6.

NUP98-Hox genes provide a proliferative advantage to bone marrow cells in vitro. (A) Expression of the fusion genes led to an increased expansion of the transduced bone marrow cells compared to that of GFP⁺ control or untransduced cells. Expansion for transduced bone marrow cells in liquid culture over 14 days and the mean (horizontal bar) of at least three experiments are shown. Expansion of cells in HOXB4- or HOXA10-transduced cells in culture ranged from 2.1 × 10³ to 3.6 × 10³ and 1.0 × 10³ to 3.1 × 10³, respectively. (B) Greater yield of day 12 CFU-S colonies after 1 week of culture of 10⁵ bone marrow cells transduced with fusions of NUP98 and Abd-B-like Hox versus NB3 and NB4 (representative experiment shown ± standard deviation [n = 3]). The P values from the t test are also shown (n = 3).

The impact on bone marrow differentiation was further investigated following a culture period of 3 to 4 weeks. The majority of cells in the GFP⁺ control cultures had undergone terminal differentiation, as evidenced by their slow growth and morphology (Fig. 7A; increased cytoplasm with increased granular content). FACS analysis (Fig. 7B) indicated a high level of expression of the c-Kit and IgE receptors, consistent with a mast cell phenotype. Under the same conditions, cultures initiated with NB4- and NB3-transduced cells consisted of a mixed population of well-differentiated and more primitive myeloid cells (Fig. 7A and B). Most strikingly, NA10- and ND13-transduced cells yielded cells almost entirely of an immature phenotype, as shown by negative expression of IgE receptors and low levels of c-Kit (Fig. 7B), blast morphology (Fig. 7A), and high proliferation rate. The greater impact on bone marrow differentiation observed for NA10 was also evidenced in vivo. Phenotypic analysis of peripheral blood of NA10- and HOXA10-transplanted mice before the onset of evident disease also showed evidence of differentiation block, with a significant (P < 0.03) decrease of circulating B220⁺ GFP⁺ B cells (Fig. 7C). In contrast, a normal distribution of lymphoid and myeloid GFP⁺ cells was observed in NB4 and HOXB4 recipients (Fig. 7C).

FIG. 7.

NUP98-Abd-B-like Hox fusion genes have a greater transforming potential in vitro. (A) Wright-Giemsa staining of transduced bone marrow cells grown for 3 weeks in liquid culture. Magnification, ×80. Note the reduced number of mature granulocytic cells for NA10- and ND13-transduced cells compared to more differentiated GFP⁺ control cells (representative experiment shown [n = 2]). (B) Immunophenotyping of cells after liquid culture. Unstained cells are in grey; IgE receptor (IgE^R) and c-Kit are shown in blue and red, respectively. (C) Lineage distribution of GFP⁺ cells in transplanted mice. Mean distribution for the cohorts ± standard deviation. Analysis was done 24 weeks posttransplant for GFP (n = 3), A10 (n = 2), and NA10 (n = 4) mice and 30 weeks posttransplant for B4 (n = 3) and NB4 (n = 3). (D) Total number of granulocyte-monocyte progenitors in serial replating. Cumulative yield is shown for the initial plating of 750 transduced bone marrow cells (representative experiment shown [n ≥ 2]).

The different transforming potential and the block in bone marrow differentiation induced by the fusion genes were further confirmed by serial replating of bone marrow clonogenic progenitor cells in methylcellulose. Control GFP⁺ cells only produced small clusters of myeloid cells (averaging 1.23 × 10³ ± 5.6 × 10² cells/colony) in secondary plating, whereas all NUP98-Hox genes produced larger secondary granulocyte-macrophage myeloid colonies (ranging from 36 × 10³ to 100 × 10³ cells/colony). Consistent with the stronger differentiation block induced by NA10 and ND13, bone marrow transduced with these genes formed large granulocyte-macrophage colonies over at least four passages (Fig. 7D). In contrast, NB4- or NB3-transduced clonogenic progenitors were essentially exhausted by the third replating (Fig. 7D; n ≥ 2).

DISCUSSION

In this study, we explored whether Hox genes have unique or overlapping or redundant roles in leukemogenesis in the context of NUP98-HOX fusion genes. Specifically, with novel NUP98-HOX fusion gene constructs, we addressed whether Abd-B-like Hox genes have common intrinsic properties that could make them prime NUP98 fusion partner candidates, or whether the required properties could extend to other Hox members that would suggest a wider overlap in Hox function. Although the engineered NUP98-Hox genes originated from two different clusters and four different paralog groups, all four fusion genes tested produced a significant overlap in effects when expressed in bone marrow cells in vitro, reminiscent of the functional redundancy commonly observed for the Hox gene family. However, not all Hox genes formed strong leukemogenic NUP98 fusion genes, but under certain circumstances, such as increased Meis1 expression, all induced myeloid leukemia.

Decisively, the Abd-B-like Hox genes formed NUP98 chimeric proteins with the greatest impact on proliferation and differentiation in vitro and leukemogenic potency in vivo, which may account for the fact that to date, only Hox genes belonging to this group have been found rearranged with NUP98 in human leukemia. Nevertheless, the leukemogenic potentials observed for NUP98-HOXB4 and HOXB4 with Meis1 support a model in which many if not all Hox genes may have the not previously appreciated capacity to contribute to leukemic transformation and point to striking overlap in properties.

Intrinsic leukemogenic potential of intact Hox genes correlates with their ability to induce AML as NUP98 fusion partners.

The novel NA10 fusion gene induced AML with kinetics similar to that of ND13, indicating that the reported leukemogenic potential of HOXA10 (37) is retained in the NA10 fusion. Our data further imply that the ability to form leukemogenic NUP98-Hox fusion genes is not restricted to the already identified Hox genes, but rather seems to be a redundant property of Abd-B-like Hox genes. These results are consistent with reported functional identity between paralogous as well as nonparalogous Abd-B-related Hox genes (10, 38, 39). Abd-B-like Hox genes are derived from a common ancestor, and therefore their homeodomains are closely related, which may indicate that their redundant function in leukemogenesis is mediated through the regulation of a similar or overlapping set of genes and/or pathways that control hematopoietic cell growth and differentiation.

Our results showing strong leukemogenic potential for NA10 but not for NB4 suggest that the intrinsic leukemogenicity of NUP98 fusions is determined primarily by the Hox fusion partner. Moreover, such intrinsic leukemogenicity appears not to be restricted to Abd-B-like Hox genes, as suggested by leukemia arising in an NB3 recipient. This differential ability of NB3 and NB4 to cause leukemia is consistent with the specific effects of their native counterparts (30, 31) and further demonstrates that closely related Hox genes are not completely functionally equivalent (26, 40).

AML is characterized by the uncontrolled proliferation of myeloid cells that accumulate at different stages where their further differentiation is blocked. The high correlation observed between the leukemogenic potential of the native and NUP98-Hox genes and their ability to increase self-renewal and block differentiation argues that the latter characteristics determine the Hox leukemogenic potential. Thus, the intrinsic ability of ND13 and NA10 and that reported for NA9 (7) to sustain increased self-renewal and strongly block bone marrow differentiation appear to be a unifying property that determines their potent leukemogenicity.

Redundant ability of Hox genes to collaborate with Meis1 to induce AML.

Despite considerable differences in the function and structure of the native and Hox-fused genes studied in this work, they were all able, though with different latencies, to cause AML in collaboration with Meis1. Importantly, our data demonstrating the capacity of NB4 and B4 to cause AML with Meis1 indicate that even Hox genes that do not cause leukemia on their own possess the redundant ability of becoming leukemogenic in the presence of this, and possibly other, cofactors.

Work from Calvo et al. showing that Meis1 may further block differentiation of myeloid progenitors overexpressing HOXA9 (5), coupled to the ability of HOXB4 to transform fibroblasts in vitro (17), could in part explain the leukemogenic activity of HOXB4 and NB4 in collaboration with Meis1. Meis1 has been shown in myeloid cells to be associated with ternary complexes containing Pbx and Hox proteins (34), in which Meis1 directly interacts with Hox proteins from paralog groups 9 to 13, or indirectly associates with paralogs 1 to 8 through Pbx (33). NA10 and NA9 both retain the Pbx interacting motifs and therefore, Meis1 may indirectly interact with these proteins through the formation of trimer complexes containing Pbx, as suggested for A9-mediated transformation of myeloid progenitors (32). Arguing against the likelihood that such direct interactions are the key to leukemic transformation, several of the fusions tested (NB3, NB4, ND13, ND13hd, and NA10hd) lacked both known Pbx1- and Meis1-interacting motifs. Nevertheless, the longer latency observed for NA10hd/Meis1 (lacking a Pbx-interacting motif) compared to NA10/Meis1 (containing a Pbx-interacting motif) suggests that interaction with Pbx proteins, though dispensable, may accelerate disease development.

How Meis1 accelerates the onset of Hox-mediated AML is still unclear. Whether Meis1 functions to increase the strength of the transcriptional activity of Hox genes on Hox-regulated promoters or works through different target genes needs to be resolved. Several non-mutually exclusive explanations could be proposed. First, in view of work from Saleh et al. showing that under some circumstances Pbx1 can repress Hox gene-mediated transcription by interacting with multiple histone deacetylases (28) and a recent report showing increased ability of HOXB4 to enhance self-renewal of HSC when Pbx1 was downregulated (18), one could propose that Meis1, known to form strong heterodimers with Pbx1, might bind and titrate off Pbx proteins. Another possibility is that the exogenous Meis1 physically interacts with endogenously expressed Hox proteins also implicated in leukemia, such as Hoxa7 and Hoxa9 (7, 25). Meis1 might also directly bind endogenous Hox-responsive genes and/or different genes, through Hox-independent activity (23). This possibility is supported by data indicating that Homothorax, a Drosophila Meis1 orthologue, possesses transcriptional activation activity of its own (12).

The NUP98-Hox fusion gene transformation models exploited here should now prove useful to dissect the molecular and cellular pathways involved in the redundant and differential leukemogenic potential of Hox genes and to provide new strategies to elucidate the role of collaborating genes.

Supplementary materials.

Primer sequences to amplify indicated Hox regions for NUP98-HOX fusion constructs were as follows: NUP98 portion, forward, 5-AGTCGGATCCTTTAACAAATCATTTGGAACACCCTT-3, and reverse, 5-TTTAAGAATTCTACTGGGGCCTGGGG-3; HOXB4 Exon2, forward, 5-CCCAGGAATTCGTAAACCCCAATTACGCC-3, and reverse, 5-ACATAAAAATGCGGCCGCTAGAGCGCGCGGGGGCC-3; HOXB3 Exon4, forward, 5-ATACCCGAATTCGCAGAGGGCTGTGGT-3, and reverse, 5-CCCATCCCCTAATCCTCGAGCGC-3; HOXA10 Exon2, forward, 5-TTATCCGAATTCGGCAATTCCAAAGGTGAA-3, and reverse, 5-TTCAATAATGCGGCCGCTCATCAGGAAAAATT-3; HOXA10 homeodomain, homeobox (hd) A10hd; forward, 5-ACTGATGAATTCAAGAGTGGTCGGAAGAAGCGC-3; reverse, 5-ATAAAAATGCGGCCGCTCACATTTTCTTCAGTTT-3; HOXD13 homeodomain, forward, 5-ACTGATGAATTCCGAAGAGGGAGGAAGAAG-3, and reverse, 5-ACTGATCTCGAGTCATTTCTTGTCCTTCACTCTTCGGT-3.