NUP98-HOXA9 Induces Long-term Proliferation and Blocks Differentiation of Primary Human CD34⁺ Hematopoietic Cells

Acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) are hematologic malignancies involving myeloid precursors (1). AML is characterized by a block in myeloid differentiation and unchecked proliferation of primitive blasts. MDS is characterized by ineffective hematopoiesis, dysplasia, and cytopenia, and often develops into AML. The nucleoporin NUP98 gene is the target of numerous chromosomal rearrangements, predominantly in AML and MDS, resulting in fusion to different partner genes, many of which belong to the homeobox family (2). NUP98-HOXA9, a prototype of the resulting fusion proteins, consists of an NH₂-terminal fragment of NUP98 fused to a COOH-terminal fragment of HOXA9. The NH₂ terminus of NUP98 includes the FG repeat domain that forms a docking site for nuclear transport carrier proteins while the COOH terminus of HOXA9 contains the DNA-binding homeodomain (2).

Investigation of the role of NUP98-HOXA9 in pathogenesis of leukemia has been carried out, to date, mostly using mouse models and cell lines. Mice transplanted with bone marrow cells that express retrovirally transduced NUP98-HOXA9 develop a preleukemic myeloproliferative disorder followed eventually by AML (3). Coexpression of another homeobox transcription factor, Meis1, or the oncogene BCR-ABL accelerates the development of AML (2). In a transgenic mouse model, another NUP98-homeobox fusion, NUP98-HOXD13, caused severe MDS followed by AML (4). These studies indicate that NUP98-HOXA9 creates a preleukemic state and that additional genetic events are required for the development of AML. In vitro, NUP98-HOXA9 promotes the growth of murine hematopoietic progenitors and blocks their differentiation (5).

The first clues to the function of NUP98-HOXA9 at the molecular level came from studies with murine cell lines in which the FG repeat region of NUP98 was shown to function as a transactivation domain when fused to a GAL4 DNA-binding domain (6). In human cells, the global effect of NUP98-HOXA9 on gene transcription was studied using a myeloid cell line, K562 (7). These studies showed that NUP98-HOXA9 acts as an aberrant transcription factor. However, its effects on proliferation, differentiation, and gene expression in primary human hematopoietic cells and the mechanisms by which it transforms these cells remain unclear.

Here we identify the effects of NUP98-HOXA9 on primary human hematopoietic progenitor/stem cells and the changes in gene expression that underlie those effects. We provide evidence that NUP98-HOXA9 has a biphasic effect on the proliferation of human CD34⁺ hematopoietic cells. An initial suppression of proliferation is associated with increased expression of IFNβ1 and marked induction of many IFN-inducible genes. This is followed by long-term proliferation, impaired differentiation, and dramatically increased numbers of primitive long-term culture-initiating cells (LTC-IC). This biphasic effect seems to recapitulate the transition of MDS to AML. Microarray analysis over a period of 16 days pointed to likely mechanisms for the biological effects of NUP98-HOXA9 by showing extensive changes in the expression of oncogenes, transcription factors, growth factors, cell cycle regulators, putative tumor suppressors, and factors involved in hematopoietic differentiation.

Retrovirus for transduction. NUP98-HOXA9 cDNA with NH₂-terminal hemagglutinin (HA) tag was subcloned into the MSCV-IRES-green fluorescent protein (GFP) retroviral expression vector as previously described (7). PG13 (CRL-10686 from American Type Culture Collection, Manassas, VA), a packaging cell line that produces GaLV-pseudotyped retrovirus, was transfected with MSCV-IRES-GFP using Lipofectamine and PLUS reagent (Invitrogen, Carlsbad, CA) and cultured 2 days in DMEM containing 10% fetal bovine serum (FBS), 2 mmol/L l-glutamine, and 100 units/mL penicillin/streptomycin. GFP-positive cells were isolated by four rounds of sorting using Elite ESP (Beckman Coulter, Fullerton, CA). Virus-containing culture supernatants were collected after 1 to 2 days at 32°C. PG13 cells producing MSCV-IRES-GFP/NUP98-HOXA9 retrovirus were kindly provided by Dr. Malcolm Moore (Sloan-Kettering Cancer Institute, New York, NY).

Culture and retroviral transduction of CD34⁺ primary cells. Frozen human CD34⁺ cells purified from mobilized peripheral blood of two patients were obtained after Institutional Review Board approval from the Bone Marrow Transplant Laboratory at Northwestern Memorial Hospital. They were meant for autologous transplantation of the now-deceased patients as a treatment for multiple myeloma. Before cell harvest, the peripheral blood of the two patients did not contain plasma cells and their bone marrows showed adequate multilineage hematopoiesis. The cells were cultured at 1 × 10⁵/mL for 2 days to preactivate in Iscove's modified Dulbecco's medium (IMDM) containing 20% FBS, 100 ng/mL Fms-related tyrosine kinase 3 (FLT-3) ligand, 20 ng/mL granulocyte/macrophage colony-stimulating factor (GM-CSF), 100 ng/mL stem cell factor, 100 ng/mL thrombopoietin, 50 ng/mL interleukin (IL)-3, 100 ng/mL IL-6 (all cytokines were from Peprotech, Rocky Hill, NJ), 2 mmol/L l-glutamine, and 100 units/mL penicillin/streptomycin (complete cytokine medium). The cells were resuspended at 0.8 × 10⁵/mL in fresh complete cytokine medium, placed in RetroNectin (Takara, Otsu, Shiga, Japan)-coated virus-preloaded plates (8), and cultured for 2 days. GFP-positive cells were isolated by sorting using a MoFlo (Dako, Glostrup, Denmark) and a portion was analyzed for expression of NUP98-HOXA9 by immunoblotting with antihemagglutinin antibody 12CA5 (Roche Applied Science, Indianapolis, IN). Another portion was cultured at 2 × 10⁵/mL in complete cytokine medium for 18 hours and collected for total RNA isolation (3 days posttransduction samples). For long-term growth, sorted cells were cultured continuously in complete cytokine medium with periodic cell counting and feeding. At 8, 10, and 16 days after transduction, cells were collected for total RNA isolation.

Flow cytometry. Antibodies (BD, Franklin Lakes, NJ) against CD11b (phycoerythrin-conjugated clone D12), CD14 (allophycocyanin-conjugated clone M5E2), CD34 (phycoerythrin-conjugated clone 581), CD45 (allophycocyanin-conjugated clone HI30 or allophycocyanin-Cy7–conjugated clone 2D1), CD71 (phycoerythrin-conjugated clone M-A712), and CD235a (allophycocyanin-conjugated clone GA-R2) were used for flow cytometric analysis. The cells were acquired with FACSCalibur (BD) or CyAn (Dako) and analyzed using the CellQuest (BD), Summit (Dako), or FCS Express (De Novo Software, Thornhill, Ontario, Canada) software.

CFC assays. Virus-transduced sorted cells were resuspended in IMDM containing 2% FBS and mixed with Methocult GF⁺ (H4435, StemCell Technologies, Vancouver, British Columbia, Canada), which consists of 1% methylcellulose, 30% FBS, 1% bovine serum albumin, 10⁻⁴ mol/L 2-mercaptoethanol, 2 mmol/L l-glutamine, 50 ng/mL stem cell factor, 20 ng/mL GM-CSF, 20 ng/mL IL-3, 20 ng/mL IL-6, 20 ng/mL granulocyte colony-stimulating factor (G-CSF), and 3 units/mL erythropoietin in IMDM. One thousand cells suspended in 1.1 mL were plated in each 35-mm dish and cultured for 14 days. Colonies were counted at ×40 magnification and classified into three categories: pure erythroid, myelomonocytic, and mixed. Cells were then suspended in IMDM containing 2% FBS and were either transferred to slides using a Cytospin centrifuge and visualized by Giemsa staining or stained with antibodies for flow cytometry.

LTC-IC assays. Ten thousand virus-transduced sorted cells were divided into two 35-mm dishes and cultured for 5 weeks with weekly one-half medium changes. The murine fibroblast cell line M2-10B4 (a gift from StemCell Technologies) that is genetically engineered to produce human G-CSF and IL-3 was used as a feeder layer after irradiation with 8,000 cGy to support the long-term culture of hematopoietic cells. The culture medium contained 12.5% horse serum, 12.5% FBS, 10⁻⁴ mol/L 2-mercaptoethanol, 2 mmol/L l-glutamine, 0.16 mmol/L I-inositol, and 16 μmol/L folic acid in MEMα (Myelocult H5100 from StemCell Technologies) with 1 μmol/L hydrocortisone. The frequency of LTC-ICs was determined by seeding varying numbers of cells in methylcellulose cultures using Methocult GF⁺ as described above and counting colonies after 17 days (9). LTC-IC numbers per 10,000 cells, the cell number originally plated at the start of the 5-week LTC-IC assay culture, were estimated by assuming an average CFC output of eight per LTC-IC (9).

Nucleofection. HA-NUP98-HOXA9 cDNA (7) was subcloned into the pTracer-CMV/Bsd vector that expresses Cycle 3-GFP-blasticidin fusion gene (Invitrogen) using EcoRI and XbaI sites. Human CD34⁺ cells preactivated as described above were resuspended at 2 × 10⁶/100 μL in Nucleofector solution (Human CD34 cell Nucleofector kit from Amaxa Biosystems, Geithersburg, MD), mixed with 5 μg DNA (HA-NUP98-HOXA9 in pTracer-CMV/Bsd or pTracer-CMV/Bsd control), and subjected to nucleofection according to the protocol of the manufacturer. The cells were then cultured at 1 × 10⁵/mL in complete cytokine medium for 3.5 hours and GFP⁺ cells were sorted using a MoFlo. Sorted cells were cultured at 2 × 10⁵/mL in complete cytokine medium for 2.5 hours and collected for antihemagglutinin immunoblotting and total RNA isolation.

RNA isolation and oligonucleotide array expression analysis. Total RNA was isolated from GFP-sorted and cultured cells using the RNeasy mini kit (Qiagen, Valencia, CA). Samples were analyzed using U133+2.0 expression array (Affymetrix, Santa Clara, CA) by the Genomic Core Laboratory of Sloan-Kettering Cancer Center as previously described (7).

Quantitative real-time PCR. Template cDNA was synthesized using SuperScript III First-Strand Synthesis System for reverse transcription-PCR (RT-PCR) (Invitrogen) from a portion of the total RNA preparation used for Affymetrix array analysis. Real-time PCR mixtures (25 μL per reaction) contained cDNA, 0.2 μmol/L each of forward and reverse primers, and 12.5 μL iQ SYBR Green Supermix (Bio-Rad, Hercules, CA). The reactions were done in triplicate in 96-well plates using the GeneAmp 5700 Sequence Detection System (Applied Biosystems, Foster City, CA) and analyzed with the software from the manufacturer. The amount of transcript was determined based on a standard curve specific for each gene and normalized to the amount of glyceraldehyde phosphate dehydrogenase transcript in the same sample.

Retrovirally transduced NUP98-HOXA9 is expressed in human CD34⁺ cells. To study the role of NUP98-HOXA9 in human leukemogenesis, human CD34⁺ hematopoietic progenitors from mobilized peripheral blood were retrovirally transduced to express the fusion protein. Flow cytometry analysis of a typical cell preparation just after thawing showed that ∼96% of the cells were CD34⁺ CD45dim⁺ progenitors. The preparation contained ∼1% CD34⁻ CD45bright⁺ lymphoid, 1% CD11b⁺ myelo/monocytic, and 9% CD235a⁺ erythroid lineage cells.

Hemagglutinin-tagged NUP98-HOXA9 was subcloned into the MSCV-IRES-GFP retroviral vector (Fig. 1A) as described (7) and GaLV-pseudotyped virus was produced in PG13 cells. CD34⁺ cells were transduced with the NUP98-HOXA9 virus or control virus prepared with empty vector. GFP-positive cells were purified by fluorescence-activated cell sorting (FACS). A typical FACS profile is shown in Fig. 1B. Antihemagglutinin immunoblotting of the sorted GFP-positive cells confirmed the specific expression of NUP98-HOXA9 protein in cells transduced with NUP98-HOXA9 virus (Fig. 1C).

NUP98-HOXA9 impairs erythroid and myeloid differentiation and induces proliferation of immature cells. The effect of NUP98-HOXA9 on cell differentiation was assessed by plating sorted cells in semisolid methylcellulose-based medium for CFC assays. After 2 weeks of culture, plates of cells transduced with NUP98-HOXA9 looked strikingly different from plates of GFP-positive control cells even without magnification: the NUP98-HOXA9 plates contained much more prominent red-colored colonies than the control plates (Fig. 2A). Although total colony numbers were comparable between the NUP98-HOXA9 and the control plates, the NUP98-HOXA9 plates contained approximately twice as many cells. Under low magnification, many of the red colonies in NUP98-HOXA9-transduced plates were large in size and several of those showed irregular contours with a mixed red/colorless morphology (Fig. 2A). In contrast, control cells showed small, tight, uniformly red erythroid colonies. Giemsa staining of cells obtained from representative red colonies showed that colonies from control cells consisted almost entirely of mature hemoglobinized erythroid precursors whereas colonies from NUP98-HOXA9-expressing cells contained a prominent population of large immature blasts admixed with maturing erythroid cells (Fig. 2A). The blasts had open chromatin, prominent nucleoli, and moderate to abundant amounts of vacuolated cytoplasm.

To examine the entire cell population, cells were recovered from the CFC plates by suspending in medium and subjected to either Giemsa staining or flow cytometry analysis. Morphologic examination of the Giemsa-stained slides revealed that cells in the NUP98-HOXA9 plates included a prominent population of blasts and consisted of higher percentages of erythroid lineage cells and lower percentages of myeloid lineage cells compared with cells in the control plates (Fig. 2B; Table 1). Furthermore, immature cells of either lineage were overrepresented among cells of the NUP98-HOXA9 plates.

Consistent with the morphologic data, flow cytometry analysis showed an increased proportion of CD235a⁺ CD45⁻ erythroid cells and a concomitant decrease of CD11b⁺ CD45⁺ myeloid cells in the NUP98-HOXA9 plates compared with control plates (Fig. 2C). The increase in CD235a⁺ erythroid cells in the NUP98-HOXA9 plates was mostly due to an increase in immature CD71bright⁺ CD235a⁺ erythroid precursors, indicating a block in erythroid differentiation. Similarly, 52% of the CD45⁺ myeloid population of the NUP98-HOXA9 plates was CD11b⁻, indicating a predominantly immature phenotype, whereas only 7.7% of the myeloid population of the control plates was CD11b⁻ (Fig. 2C). In summary, these results show that NUP98 inhibits myeloid and erythroid differentiation while increasing the numbers of immature cells.

NUP98-HOXA9 has a biphasic effect on CD34⁺ cell proliferation. To determine the effect of NUP98-HOXA9 on the proliferation of human hematopoietic cells, sorted NUP98-HOXA9-transduced human CD34⁺ cells and control cells were continually grown in liquid culture in the presence of cytokines with periodic cell counting. Typical growth patterns observed in one of three independent experiments are shown in Fig. 3. The growth of NUP98-HOXA9 cells was slower than that of control cells in the early stages of the culture (Fig. 3A). However, by day 9, this trend was reversed in all three experiments and NUP98-HOXA9 cells grew faster than control cells. By day 12, the cumulative number of NUP98-HOXA9 cells exceeded that of control cells. NUP98-HOXA9 cells continued to proliferate for an average of 54.3 ± 1.5 days; after which, their numbers slowly declined. Control cells stopped growing significantly earlier, after an average of 27.3 ± 1.2 days, and their numbers were 2 orders of magnitude lower at their peak (Fig. 3B). These data suggest that after an initial period of growth suppression, NUP98-HOXA9 induces the proliferation of primitive cells that are capable of long-term growth.

NUP98-HOXA9 promotes proliferation of LTC-ICs. The vast majority of CD34⁺ cells are progenitors with no long-term self-renewal capabilities. On the other hand, a small minority of more primitive progenitors/stem cells, known as LTC-ICs, are capable of self-renewal and remain clonogenic after prolonged in vitro culture. These are the most primitive hematopoietic cells that can be assayed in vitro (10). To determine the effect of NUP98-HOXA9 on the number of LTC-ICs, virus-transduced and sorted cells were cultured on a feeder layer of irradiated M2-10B4 cells (9) for 5 weeks with weekly one-half medium changes. Cells recovered after 5 weeks were plated for CFC assays and colonies were counted 17 days later. Cells transduced with NUP98-HOXA9 showed a drastic increase in the number of colonies in four separate experiments using cells from two unrelated individuals. The data are summarized in Table 2 expressed as the average numbers of LTC-ICs per 10,000 original input cells.

Extensive changes in gene expression underlie the biological effects of NUP98-HOXA9. The results described above show that NUP98-HOXA9 inhibits hematopoietic differentiation, has a biphasic effect on proliferation, and induces a dramatic increase in the numbers of primitive self-renewing cells. To identify the changes in gene expression that underlie these effects, a time-course microarray study was conducted. Cells were retrovirally transduced with NUP98-HOXA9 or control virus and sorted for GFP expression as described above. RNA was isolated 3, 8, 10, and 16 days after transduction and subjected to oligonucleotide array expression analysis. Only those genes that exhibited ≥2-fold difference in their expression between the NUP98-HOXA9 cells and the controls were judged to be significantly affected by NUP98-HOXA9. Among the numerous genes significantly affected by NUP98-HOXA9, representative genes that are most likely to explain its effects on cell proliferation, differentiation, and transformation are listed in Table 3. The supplemental list of all affected genes is available online (Supplementary Tables S1-S5).

The most striking feature of the gene profiling results was the dramatic up-regulation by NUP98-HOXA9 of a large number of IFN-inducible genes in the 3-day samples (Table 3). Most of these genes belong to the IFNα/β pathways while some belong to the IFNγ pathway. IFNs are known to suppress hematopoietic cell growth (11). Thus, activation of the IFN signaling cascade may explain the early decrease in proliferation seen in NUP98-HOXA9-transduced cells. Similarly, cyclin-dependent kinase inhibitor 1C (CDKN1C; also known as p57, Kip2) is known to play a role in cell cycle arrest in hematopoietic cells (12) and its early induction (Table 3) may play a role in the early proliferation block in response to NUP98-HOXA9. The expression of both IFN-inducible genes and CDKNC1 declined or returned to baseline later on as NUP98-HOXA9-transduced cells entered the proliferative phase.

While expression of the IFN-inducible genes peaked on day 3, many genes associated with cell proliferation and oncogenesis peaked at day 8 or beyond (Table 3), approximately coinciding with the beginning of the proliferative phase of NUP98-HOXA9-expressing cells (Fig. 3). These include a number of homeobox transcription factors, growth factors, receptor tyrosine kinases, and other oncogenes. Among the induced genes were several that are known to be involved in myeloid leukemogenesis, including HOXA9, HOXA7, MEIS1, FLT3, KIT, WT1, and ANGPT1 (see Discussion). In addition, several genes with putative tumor suppressor functions were down-regulated, especially at later time points. At the cell cycle level, the increased cell proliferation was accompanied by up-regulation of cyclin A1, a cell cycle regulator that is overexpressed in AML (13), and down-regulation of several cell cycle inhibitors (Table 3).

Microarray results (Table 3) confirmed the inhibition of myelomonocytic differentiation by NUP98-HOXA9 and suggested possible mechanisms. Genes encoding secondary neutrophil granule proteins such as lactoferrin, transcobalamin 1, and ABP1, which are expressed later in myeloid differentiation (14), were down-regulated. On the other hand, some genes encoding primary granule proteins, such as neutrophil elastase and cathepsin G (14), were up-regulated, whereas others, such as BPI and cathepsin B (14), were down-regulated. These findings are consistent with a partial differentiation block at around the blast/promyelocyte stage (14). Indeed, most cases of AML associated with NUP98 gene rearrangements belong to subcategories M2, M4, and M5 in which myeloid differentiation is only partially blocked (2). Interestingly, CCAAT/enhancer binding protein ε (C/EBPε), a transcription factor that is essential for terminal granulocytic differentiation and is important for the expression of secondary granule proteins (15), was down-regulated by NUP98-HOXA9 (Table 3). Further, the down-regulation of C/EBPε was preceded by down-regulation of retinoic acid receptor α (RARα), a transcription factor known to promote the expression of C/EBPε (16). Thus, the inhibition of myeloid differentiation by NUP98-HOXA9 may involve down-regulation of RARα leading to down-regulation of C/EBPε, which results in down-regulation of granule proteins and other markers of terminal myeloid differentiation.

The microarray data also point to a block in megakaryocytopoiesis. For example, pro-platelet basic protein (PPBP; β-thromboglobulin), a component of platelet α granules (17), and FLI1, an Ets family transcription factor that drives megakaryocytopoiesis (18), were down-regulated whereas the megakaryocytopoiesis inhibitor thrombospondin 1 (19) was up-regulated (Table 3).

Most hemoglobin genes were up-regulated, consistent with the increased numbers of erythroid precursors observed in CFC cultures of cells transduced with NUP98-HOXA9 (Table 1). Some genes associated with erythroid differentiation, such as APOE, AQP1, and ETS1 (20–22), were down-regulated. As globin genes are expressed early during erythroid differentiation (23), these data are consistent with the block in erythroid maturation observed in NUP98-HOXA9-transduced cells by morphology and flow cytometry (Fig. 2B and C).

To verify the microarray results, RNA levels of selected genes were determined by quantitative RT-PCR. The fold changes in gene expression (shown between brackets in Table 3) were similar to those obtained by microarray analysis.

Early NUP98-HOXA9 target genes include IFNβ1 and many transcription factors. NUP98-HOXA9 acts as an aberrant leukemogenic transcription factor that probably binds DNA through its HOXA9 homeodomain (6, 7), yet its direct target genes in primary human hematopoietic cells are not known. While the gene changes observed 3 days or more after NUP98-HOXA9 transduction yielded important information relevant to its biological effects (Table 3), many of these genes are probably affected by NUP98-HOXA9 indirectly, rather than being direct transcriptional targets of NUP98-HOXA9. In addition, the mechanism of the strong induction of IFN-inducible genes by NUP98-HOXA9 at 3 days posttransduction was not clear. To address these issues, a microarray analysis needed to be done within hours of NUP98-HOXA9 expression in cells, which is not possible with retroviral transduction. We therefore used the Amaxa Nucleofector device that introduces DNA into primary cells efficiently and allows rapid expression of the protein of interest. CD34⁺ cells were transfected with a NUP98-HOXA9 construct that expresses a GFP marker from a separate promoter; control cells were transfected with the empty vector. GFP-positive cells were purified by FACS after 3.5 hours, followed by further culture for 2.5 hours. Expression of NUP98-HOXA9 protein was confirmed by immunoblotting (not shown) and total RNA was isolated. Microarray analysis showed that NUP98-HOXA9 up-regulated IFNβ1 expression within 6 hours posttransfection (Table 3). In contrast, all but one of the IFN-inducible genes that were up-regulated after 3 days were still at baseline at this time point. These results indicate that the early induction of IFNβ1 expression led to the subsequent induction of the genes involved in IFN-signaling pathways. Overexpression of IFNβ1 was confirmed by quantitative real-time PCR (Table 3).

Another interesting observation is that many transcription factors, particularly homeobox factors, were already induced by NUP98-HOXA9 at the 6-hour time point. This suggests that the effects of NUP98-HOXA9 on CD34⁺ cells may be mediated, at least in part, by the induction of other transcription factors, especially those of the homeobox family.

In the present study, we undertook a comprehensive investigation of the biological effects of NUP98-HOXA9 on primary human CD34⁺ progenitor/stem cells and correlated those effects with changes in global gene expression. These studies followed the effects of NUP98-HOXA9 from the earliest time point within 6 hours of the introduction of the oncogene up to several weeks thereafter. The results provide insights into the mechanisms by which NUP98-HOXA9 transforms human hematopoietic cells and provide robust assays for the further dissection of those mechanisms and for preliminary testing of potential therapies.

Continuous culture of NUP98-HOXA9-transduced CD34⁺ cells in liquid medium revealed initial growth inhibition (Fig. 3). Because the vast majority of CD34⁺ cells are lineage-committed progenitors (24), the cells that failed to grow and/or died at the initial stage consist mostly, if not entirely, of these progenitors. IFNs are known for their growth-inhibitory effects on cells, including those of the hematopoietic lineage (11). Therefore, the early induction of IFNβ1 and IFN-inducible genes (Table 3) may explain the antiproliferative effect of NUP98-HOXA9 in the first days after transduction. Interestingly, the oncoprotein P210/BCR-ABL has also been reported to up-regulate a group of IFN-responsive genes (25). In addition, IFN-inducible genes are up-regulated in preleukemic conditions such as aplastic anemia, paroxysmal nocturnal hemoglobinuria, and MDS (26). Thus, induction of IFN pathways may reflect a common response during leukemogenesis. The mechanism by which NUP98-HOXA9 up-regulates IFNβ1 expression remains to be determined, and it is not clear whether the IFN response is part of the transformation process or a reaction against it.

The initial suppression of proliferation by NUP98-HOXA9 was followed by several weeks of sustained proliferation (Fig. 3), suggesting an expansion of primitive progenitors and/or stem cells. This notion was confirmed by the dramatic increase in the numbers of LTC-ICs (Table 2). This biphasic effect of NUP98-HOXA9 on hematopoietic cells, with early growth inhibition and late proliferation of primitive cells, is reminiscent of the progression of MDS to AML. Indeed, NUP98-HOXA9 and other NUP98 fusions have been observed in cases of MDS (2), and mice transgenic for NUP98-HOXD13, another leukemogenic NUP98-HOX fusion, develop severe MDS followed by acute leukemia (4). Interestingly, NUP98-HOXA9 induced proliferation of erythroid precursors (Table 1; Fig. 2), which is often a feature of human MDS (27). In addition, erythroid proliferation has been reported when leukemogenic proteins such as BCR-ABL, constitutively active STAT5A, and AML1-ETO are overexpressed in human CD34⁺ cells (28–30). Similar to NUP98-HOXA9, AML1-ETO caused initial inhibition of cell proliferation followed by increased long-term proliferation and expansion of primitive cells (30–32). However, the initial growth inhibition lasted ∼4 weeks with AML1-ETO as opposed to only 9 days with NUP98-HOXA9 (Fig. 3). In addition, human CD34⁺ cells transduced with AML1-ETO continued to grow for over 7 months in culture whereas cells transduced with NUP98-HOXA9 stopped growing after ∼54 days. It remains to be determined whether NUP98-HOXA9 increases the number of LTC-ICs by stimulating the proliferation of multipotent progenitors/stem cells or whether it endows more mature progenitors with self-renewal capacity.

Microarray analysis revealed marked changes in gene expression that point to the mechanisms behind the long-term proliferation of cells transformed by NUP98-HOXA9. The IFN response, which peaks at day 3, began to subside at day 8 as evidenced by dramatic decreases in the levels of IFN-responsive genes. This coincided with the beginning of the proliferative phase of NUP98-HOXA9-transduced cells at around day 9 (Fig. 3). In contrast to IFN pathway genes, the expression of most of the genes associated with proliferation and oncogenesis peaked at later time points (8 days or more after transduction). As discussed below, several of these genes are likely to play an important role in the pathogenesis of AML in patients with NUP98-HOXA9 gene rearrangements.

Homeobox genes. These encode a group of transcription factors well known for their involvement in normal hematopoiesis and leukemogenesis, including HOXA9, HOXA7, HOXA4, MEIS1, PBX3, and others (Table 3). Several homeobox transcription factors are expressed in primitive hematopoietic cells and down-regulated with differentiation (33). HOXA9 plays an important role in hematopoietic stem cell proliferation (34) and its overexpression in mouse bone marrow causes AML, which is accelerated by coexpression of MEIS1 (3). MEIS1 also accelerates the induction of AML by NUP98-HOXA9 in mice (3). Several homeobox genes, including HOXA9, HOXA7, and MEIS1, are frequently overexpressed in human AML and are associated with worse prognosis (35–39). Interestingly, a number of homeobox genes, particularly HOXA7 and HOXA9, are downstream targets of leukemogenic fusions of the MLL gene at chromosome 11q23 and are thought to mediate the leukemogenic effects of these fusions in AML (40). The early and sustained induction of homeobox transcription factors by NUP98-HOXA9 (Table 3) suggests that they may play a similar role in mediating leukemogenesis by NUP98 fusions, and their overexpression may contribute to the worse prognosis of patients with these fusions.

Other leukemogenic transcription factors. Several other transcription factors with a strong association with AML were up-regulated by NUP98-HOXA9. These include EVI1, MEF2C, SOX4, and WT1 (Table 3). EVI1 is overexpressed in some cases of AML in which it is associated with a poor prognosis (41). Its overexpression in response to NUP98-HOXA9 may contribute to the poor prognosis in patients with NUP98 gene rearrangements. Interestingly, EVI1 is thought to counteract the antiproliferative effects of IFNα (42) and may therefore contribute to the proliferation of NUP98-HOXA9-expressing cells in spite of the activation of IFN pathways. SOX4 can cooperate with either MEF2C or EVI1 to induce AML in mice (43, 44). WT1 plays a key role in cell proliferation in many human neoplasms, particularly acute leukemia, and is used to monitor minimal residual disease AML and MDS (45).

Receptor tyrosine kinases. In addition to aberrant transcriptional regulation, the pathogenesis of AML often involves abnormalities of signaling pathways. Two receptor tyrosine kinases well known for their role in the pathogenesis of AML were up-regulated by NUP98-HOXA9: FLT3 kinase and KIT (Table 3). Activation of FLT3 by mutation and/or internal tandem duplication is observed in about one third of AML cases and increased expression of wild-type FLT3 is seen in some AML patients in which it seems to be associated with a worse prognosis (39, 46). Similarly, KIT, the receptor for stem cell factor, is expressed in blasts from 80% of AML patients and is activated by mutation in many cases of AML (46).

Growth factors. Several secreted factors associated with cell proliferation and leukemogenesis were up-regulated by NUP98-HOXA9, including renin, TSLP, angiopoietins, and fibroblast growth factor 18 (FGF18) (Table 3). Among them, the induction of renin expression was particularly strong. Renin is a constituent of the renin-angiotensin system (RAS) and converts angiotensinogen to angiotensin I, which then is converted by angiotensin-converting enzyme to angiotensin II, a major regulator of blood pressure and electrolyte balance. In addition to its classic role, the RAS may play an important role in hematopoiesis (47). Indeed, angiotensin II has been shown to increase the proliferation of both primitive and committed hematopoietic progenitors. Blasts from many AML patients express renin whereas normal bone marrow cells do not. Thus, it is possible that the enhanced expression of renin is a part of the mechanism for uncontrolled cell growth in some leukemias.

Examination of changes in gene expression 6 hours after introduction of NUP98-HOXA9 into CD34⁺ cells provided a list of likely direct transcriptional targets of NUP98-HOXA9 (Table 3 and Supplementary Table S1). In addition, it provided clues to the earliest events in NUP98-HOXA9 oncogenesis. The leukemogenic genes induced by NUP98-HOXA9 within 6 hours are mostly transcription factors, including members of the homeobox family and EVI1 (Table 3). This suggests that the initiating oncogenic event is an abnormal transcriptional program that leads to secondary changes in gene expression, culminating in blocked differentiation and increased proliferation. For example, NUP98-HOXA9 induces renin at 3 days but not at 6 hours. Because the renin gene promoter contains HOX/PBX-binding sites and is thought to be regulated by HOX/PBX proteins (48), its induction by NUP98-HOXA9 may be mediated by the early induction of homeobox transcription factors.

Interestingly, the leukemogenic receptor tyrosine kinase FLT3 is also up-regulated within 6 hours of NUP98-HOXA9 transduction. Constitutively active FLT3 can cooperate with leukemogenic fusion transcription factors such as PML-RARα and AML1-ETO in inducing leukemia in mice (49, 50). Therefore, early overexpression of FLT3 may similarly cooperate with NUP98-HOXA9 in the leukemic transformation of human cells.

In summary, the present data show that NUP98-HOXA9 alters gene transcription in human hematopoietic precursors, resulting in disruption of cell differentiation and a biphasic effect on cell growth. The observed changes in gene expression identify likely direct target genes for NUP98-HOXA9 and provide clues to the mechanisms by which it transforms human hematopoietic cells. The assays described, particularly the LTC-IC assay, can be used to further dissect these mechanisms and to test potential therapeutic agents.

Grant support: NIH grant RO1 HL 82549 and Myeloid Malignancies Specialized Center of Research grant from the Leukemia and Lymphoma Society.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Malcolm Moore for retrovirus-producing cells and Dr. Richard Burt and Marcelo Villa for help with obtaining CD34⁺ primary human hematopoietic cells.