Expression of a CALM-AF10 Fusion Gene Leads to Hoxa Cluster Overexpression and Acute Leukemia in Transgenic Mice

To assess the role of the CALM-AF10 fusion gene in leukemic transformation in vivo, we generated transgenic mice that expressed a CALM-AF10 fusion gene. Depending on the transgenic line, at least 40% to 50% of the F₁ generation mice developed acute leukemia at a median age of 12 months. Leukemic mice typically had enlarged spleens, invasion of parenchymal organs with malignant cells, and tumors with myeloid markers such as myeloperoxidase, Mac1, and Gr1. Although most leukemias were acute myeloid leukemia, many showed lymphoid features, such as CD3 staining, or clonal Tcrb or Igh gene rearrangements. Mice were clinically healthy for the first 9 months of life and had normal peripheral blood hemograms but showed impaired thymocyte differentiation, manifested by decreased CD4⁺/CD8⁺ cells and increased immature CD4⁻/CD8⁻ cells in the thymus. Hematopoietic tissues from both clinically healthy and leukemic CALM-AF10 mice showed up-regulation of Hoxa cluster genes, suggesting a potential mechanism for the impaired differentiation. The long latency period and incomplete penetrance suggest that additional genetic events are needed to complement the CALM-AF10 transgene and complete the process of leukemic transformation. [Cancer Res 2007;67(17):8022–31]

The analysis of fusion genes produced by nonrandom, recurring chromosomal translocations has proved to be a rich source of insights into the process of leukemic transformation (1–3). The recurring t(10;11)(p14;q21) chromosomal translocation has been recognized in patients with a wide array of hematologic malignancies,⁴

most commonly pre–T-lymphoblastic leukemia/lymphoma (pre–T-LBL) and acute myeloid leukemia (AML). A recent analysis of pre–T-LBL patients with the t(10;11)(p14;q21) showed that the vast majority of samples either expressed the T-cell receptor (TCR) γδ heterodimer or were immature, undifferentiated T cells that expressed no TCR (4). Moreover, most of the AML or acute undifferentiated leukemia patients in that series showed TCR gene rearrangements, suggesting that the transformed cell may have possessed the potential for lymphoid differentiation (4). The t(10;11)(p14;q21) translocation breakpoint was initially cloned from the U937 cell line, a cell line with monocytic features that was initially established from a patient with histiocytic lymphoma and was shown to result in an in-frame fusion between the CALM and AF10 genes (5).

The CALM (for Clathrin Assembly Lymphoid Myeloid; also known as PICALM) gene is ubiquitously expressed and encodes a 652-amino-acid residue protein (5). CALM has structural similarity to the clathrin-binding protein AP180 (6) and contains an epsin NH₂-terminal homology domain, DPF (ASP-Pro-Phe) and NPF (Asn-Pro-Phe) motifs, and clathrin-binding sequences (6, 7). The CALM protein is normally involved in endocytosis and the formation of clathrin-coated vesicles from cell membranes. Mutations in the CALM gene lead to defective endocytosis and are responsible for the iron metabolism defects seen in fit1 mice (7).

AF10 (also known as MLLT10) was initially cloned as an MLL chromosomal translocation partner gene in patients with AML (8). Similar to CALM, AF10 is ubiquitously expressed and encodes a 1,027-amino-acid protein. AF10 contains NH₂-terminal zinc-fingers, a COOH-terminal leucine zipper, and an AT-hook domain. AF10 is thought to normally function as a transcription factor and has been shown to be leukemogenic when fused to MLL (9). Deletion analysis of the MLL-AF10 fusion protein has shown that the AF10 leucine zipper motif is required for leukemic transformation (9).

The CALM-AF10 fusion genes retain all but the final four amino acid residues of the CALM gene, which is fused to the COOH-terminal portion of the protein encoded by the AF10 gene (4, 5). Although the CALM fusion point is constant, at least four different AF10 fusion points have been described; these have been classified as 5′ or 3′ transcripts, depending on how far 5′ on the AF10 sequence the fusion occurs (4). There is a correlation between the fusion point and leukemic phenotype, with 3′ fusion transcripts being associated with more immature T cells that do not express a TCR on the cell surface and 5′ fusion transcripts being associated with TCRγδ T cells (4).

Because cancer develops in the context of an intact organism, with the influence of nearby cells and tissues, genetically engineered animal models have become a standard model for showing that a lesion associated with any form of malignancy causes the malignancy (10). Moreover, generation of mouse models allows one to study the process of malignant transformation over time in vivo. In this report, we show that expression of CALM-AF10 in murine hematopoietic cells leads to overexpression of Hoxa cluster genes, impairs hematopoietic differentiation, and is ultimately leukemogenic.

Generation of transgenic mice. A CALM-AF10 fusion cDNA fragment was amplified from the U937 cell line by using reverse transcription-PCR (RT-PCR), and fragments of CALM and AF10 derived from EST clones were ligated to the 5′ and 3′ portions of the fusion cDNA. A human β-globin 5′ untranslated region (UTR) was ligated to the 5′ portion of a full-length CALM-AF10 cDNA, and the resultant cDNA was cloned into the SfiI and NotI sites of the HS21/45-vav vector (11). To delete all plasmid backbone sequences, the CALM-AF10 insert was then shuttled into the pSV40zeo (Invitrogen) vector by using a partial HindIII digest. The resultant plasmid, pSVZvavCA, was digested with PmeI to remove all plasmid sequences. The CALM-AF10 expression cassette was purified over a sucrose gradient, and FVB/N zygotes were microinjected with the construct at the National Cancer Institute transgenic animal facility. Transgenic mice that had incorporated the construct were identified by Southern blot analysis of tail DNA. The transgenic lines were maintained by breeding to wild-type (WT) FVB/N mice. Offspring of the founders were genotyped by PCR amplification of the CALM-AF10 fusion gene from tail biopsy using primers 5′-TGTTCCTGTAATGACGCAACCAACC-3′ and 5′-CTCTGGAATATACAGGGCACAAACA-3′ and a thermal cycling profile of 94°C for 3 min; 34 cycles of 94°C for 30 s, 62°C for 30 s, 72°C for 1 min; followed by a terminal extension of 72°C for 10 min.

Expression of the CALM-AF10 transgene. Expression of the CALM-AF10 transgene was determined by Northern blot analysis and RT-PCR. Total RNA was isolated from thymus, spleen, bone marrow, liver, and kidney of mice using Trizol (Invitrogen) reagents and protocols. Total RNA (10 μg) was size fractionated on an agarose/formaldehyde gel, transferred to a nitrocellulose membrane as previously described (12), and hybridized with a ³²P-labeled human CALM probe (0.5-kb HindIII-EcoRI fragment; nucleotides 1,588–2,121 of NM_007166.2). The probe was labeled with [³²P]dCTP using Ready-To-Go DNA Labeled Beads (dCTP; Amersham Bioscience) and following the manufacturer's recommended protocol. Stringency washes were 20 min × 2 with 0.1× SSC/0.1% SDS at 52°C. RT-PCR amplification of the CALM-AF10 mRNA was done by reverse transcription of 1-μg total DNase-treated (Ambion) RNA using Superscript II reverse transcriptase and random hexamer primers (Invitrogen) in a volume of 20 μL. One to two microliters of the reverse transcription reaction were used to amplify the CALM-AF10 fusion with primers 5′-TGTTCCTGTAATGACGCAACCAACC-3′ and 5′-CTCTGGAATATACAGGGCACAAACA-3′ and a thermal cycling profile of 94°C for 3 min; 34 cycles of 94°C for 30 s, 62°C for 30 s, 72°C for 1 min; followed by a terminal extension of 72°C for 10 min.

Evaluation of leukemic and healthy mice. A cohort of 38 transgenic and 33 WT littermate controls were housed together and observed daily for signs and symptoms of disease. Statistical analysis was done by χ² with 1 degree of freedom at 18 months of age. Whole blood was obtained from tail veins for complete blood counts (CBC) and morphologic evaluation in an attempt to detect leukemia in clinically healthy mice. Mice that showed signs of disease such as ruffled fur, hunched posture, or difficulty in breathing were euthanized for postmortem evaluation. Mice that were found dead were also dissected and tissues placed in formalin as described below, unless the tissues were judged to be severely decomposed. Tissue samples including the liver, kidney, spleen, thymus, bone marrow, lung, and heart were fixed in 10% neutral buffered formalin (Sigma), paraffin embedded, sectioned at 5 μm, and stained with H&E or anti-myeloperoxidase (MPO; DAKO), CD3 (DAKO), F4/80 (Caltag), and B220 (CD45R; PharMingen). The Bethesda proposals for classifying nonlymphoid hematopoietic and lymphoid neoplasms in mice were used as guides to evaluate tissues from CALM-AF10 mice (13, 14). Single-cell suspensions prepared from thymus, spleen, and/or bone marrow were incubated with FITC-conjugated antimouse CD8, B220, and Gr-1; phycoerythrin-conjugated antimouse CD4, IgM, and Mac-1 (PharMingen); allophycocyanin-conjugated antimouse CD19, CD117, and IgM (eBioscience); and phycoerythrin-Cy5–conjugated antimouse CD24, and then analyzed by three-color flow cytometry to determine immunophenotype. To identify Lin⁻/Sca-1⁺/c-Kit⁺ (“LSK”) cells, mouse bone marrow cells were resuspended in HBSS containing 2% fetal bovine serum (HF2) to 1 × 10⁷/mL. The cells were incubated with biotin-conjugated lineage antibody cocktail (CD5, TER119, B220, GR-1, Mac-1, 7-4; Stemcell Technologies), allophycocyanin-conjugated streptavidin (PharMingen), FITC-conjugated anti–c-Kit (PharMingen), and phycoerythrin-conjugated anti–Sca-1 (PharMingen). The stained cells were resuspended in HF2 containing 1 μg/mL propidum iodide (Sigma). Four-color staining was used to identify LSK cells. Statistical differences between groups were analyzed with two-sided Student's t test.

Southern blot analysis for Tcrb, Tcrd, and Igh gene rearrangements. Genomic DNA from mouse liver, spleen, or thymus was digested with either HindIII or SstI (for Tcrb and Tcrd gene rearrangements) or either EcoRI or XbaI (for Igh rearrangements), size fractionated on a 0.8% agarose gels, and transferred to nitrocellulose membranes, as previously described (15). The nitrocellulose membrane was hybridized to a ³²P-labeled TCRB probe that detects the constant region of both Tcrb1 and Tcrb2 genes, to a ³²P-labeled Jδ1 TCRD probe (16) that detects the Jδ1 and constant regions of Tcrd, or to a ³²P-labeled murine Igh probe that hybridizes to the JH3-JH4 region of the mouse Igh locus (17).

Real-time RT-PCR assay for Hoxa5, Hoxa7, Hoxa9, Hoxa10, Hoxa11, Hoxa13, Hoxb4, Hoxd13, and Meis1. Total RNA (1 μg) from transgenic or WT spleen, bone marrow, and thymus, as well as spleen infiltrated by leukemic cells, was reverse transcribed as described above. Real-time RT-PCR was done on a 7500 Fast Real-time TaqMan PCR system (Applied Biosystems) using aliquots of first-strand cDNA as templates for mouse Hoxa5, Hoxa7, Hoxa9, Hoxa10, Hoxa11, Hoxa13, Hoxb4, Hoxd13, and Meis1 with Applied Biosystems primer and probe sets. Primer details are available on request. The expression of the 18S rRNA was used as an endogenous control. All reactions were done in triplicate and the −ΔΔC_T mean and SE were calculated for each sample. Values for the transcript of interest were normalized to the 18S rRNA value and compared with the expression in WT bone marrow.

To generate genetically identical mice that expressed a CALM-AF10 fusion gene, we used RT-PCR to amplify a portion of the CALM-AF10 fusion cDNA from the U937 cell line. Although several different CALM-AF10 fusion cDNAs have been described (4), the predominant fusion cDNA that we detected in the U937 cell line joined CALM nucleotide 2,230 (GenBank reference NM_007166.2) to AF10 nucleotide 241 (GenBank reference NM_001009569.1). We then extended the fusion cDNA 5′ and 3′ by ligating portions of CALM and AF10 cDNAs to the CALM-AF10 PCR product and introduced a human β-globin 5′ UTR to aid in the translation of the CALM-AF10 fusion mRNA. The entire coding sequence of the CALM-AF10 cDNA was 4,881 nucleotides encoding a 1,627-amino-acid protein (see Supplementary Fig. S1 for nucleotide sequence). This cDNA was cloned into the HS21/45-vav vector (11) that uses 5′ and 3′ vav regulatory elements to direct expression of cDNA inserts specifically in hematopoietic tissues (Fig. 1A).

Transgenic CALM-AF10 mice were generated by pronuclear injection of FVB/N single-cell embryos, and eight potential founders were identified by Southern blotting of tail DNA. Four founders were bred and all four transmitted the transgene. F₁ mice were euthanized to determine expression of the transgene. As expected, the CALM-AF10 transgene was expressed in hematopoietic tissues (thymus, bone marrow, and spleen) but not in nonhematopoietic tissues (liver and kidney; Fig. 1B and C). CBCs were obtained from clinically healthy offspring of the CALM-AF10 founders ages 6 to 10 months; although the CALM-AF10 transgenic mice had increased numbers of atypical lymphocytes (data not shown), there were no statistical differences in WBC, neutrophil, lymphocyte, hemoglobin, or platelet counts (Supplementary Table S1).

CALM-AF10 mice develop acute leukemia. A cohort of transgenic offspring from two founders, C10 and E6, were followed for 18 months and were euthanized when signs and symptoms of leukemia, such as tachypnea, lethargy, ruffled fur, hunched posture, or lymphadenopathy, were detected. Additionally, mice that were found dead in their cage were necropsied when possible and tissues harvested for histology. As shown in Fig. 1D, 83% of the E6 transgenic mice and 70% of the C10 transgenic mice died by 18 months of age. At least 9 of the 18 (50%) E6 mice and 8 of the 20 (40%) C10 mice had clear signs of leukemia; several additional mice were found dead but were too autolytic to be analyzed. WT mice that died during the study showed no evidence of leukemia. In addition to the E6 and C10 offspring, one potential founder mouse, D5, developed a myeloid leukemia, and one of six offspring of a fourth founder (D3) died of a myeloid leukemia.

Leukemic mice typically had hunched posture, ruffled fur, and tachypnea. Gross examination revealed marked splenomegaly and hepatomegaly, with prominent scattered white foci. Histologically, the splenic red pulp was effaced by an expanding myeloid infiltrate accompanied by follicular atrophy. Peripheral blood examination showed circulating blasts with a high nuclear/cytoplasmic ratio, and the bone marrow was replaced by cells with a similar appearance (Fig. 2). In addition to the spleen, parenchymal organs such as liver, lung, kidney, and brain were infiltrated with malignant cells. Involved tissues were evaluated with immunohistochemical stains including the myeloid marker MPO, the B-cell marker B220, the T-cell marker CD3, and the monocytic marker F4/80 (Table 1). CBCs, fluorescence-activated cell sorting (FACS), and antigen-receptor gene rearrangements were also analyzed on a subset of leukemic mice.

As shown in Table 1, all 17 of the analyzable leukemias were MPO positive, indicating that these mice had myeloid leukemia. Three additional mice (7004, 7092, and 7061) were found dead and showed leukemic infiltration of parenchymal organs, including the liver, kidney, and spleen, but did not stain positively for MPO, B220, CD3, or F4/80, raising the possibility that these mice had undifferentiated leukemias. However, given the phenotype of the other leukemic mice, it is likely that these were also myeloid leukemias but that the MPO antibody was nonreactive because the tissues were partially autolytic. Nine of these analyzable leukemias were also B220⁺, and FACS analysis done on a subset of these mice showed a population of leukemic cells that were Mac1⁺/B220⁺ (Fig. 2). Eight of the MPO⁺ leukemias were negative for B220; a subset of these were analyzed by FACS and shown to be Mac1⁺/B220⁻ (Supplementary Fig. S2). Two mice (7026 and 2953) developed myelomonocytic leukemia as evidenced by both MPO and F4/80 immunohistochemical staining; one of these also stained positive for CD3 (Supplementary Fig. S3). Taken together, these findings show that expression of a CALM-AF10 fusion gene leads to an acute leukemia, with a long latency period and incomplete penetrance.

Immunophenoptype analysis. We observed that more than half of the leukemic mice had either MPO⁺/B220⁺ cells within tumor infiltrates or were Mac1⁺/B220⁺ by FACS analysis. We considered the possibility that these B220⁺ leukemic cells might have other properties of B cells, or that they may be more immature biphenotypic cells, suggesting that the leukemia had arisen in a common early progenitor cell with the potential to differentiate along the myeloid or lymphoid lineages. Therefore, the leukemic Mac1⁺/B220⁺ population was assayed for the expression of c-Kit (CD117), CD19, CD24, or IgM. As shown in Fig. 3, although the leukemias were consistently positive for Mac1, the B220 staining was quite variable (Fig. 3, compare 9001 and 7160). The Mac1⁺/B220⁺ cells were typically positive for CD24, negative for IgM and CD19, but showed variable expression of CD117, ranging from 23% to 72% positive cells. Of note, some of the mice (such as 9001; see Fig. 3A) had a distinct B220⁺ population that was distinct from the Mac1⁺/B220⁺ population (Fig. 3A,, gate R4). This “bright” B220⁺ population was not present in bone marrow from the leukemic mouse (data not shown), and these cells were positive for CD19 and IgM (not shown), leading us to conclude that the Mac1⁻/B220 bright population in the spleen represented contaminating normal B cells. The B220 “dim” population (Fig. 3A , gate R5) from these same mice was negative for both IgM and CD19 (not shown).

Antigen receptor gene rearrangements. Because some of the leukemic mice stained for either the B-cell marker B220 or the T-cell marker CD3 in addition to myeloid markers, we considered the possibility that some of these cells might have additional features of T or B lymphocytes, such as clonal Tcrb, Tcrd, or Igh gene rearrangements. As shown in Fig. 4, some of these leukemias had clonal T-cell and/or immunoglobulin gene rearrangements; indeed, one mouse (2953) had clonal rearrangements of Tcrb, Tcrd, and Igh. In all, 3 of 15 mice analyzed had clonal Tcrb gene rearrangements, 1 of 15 had a clonal Tcrd gene rearrangement, and 8 of 15 mice analyzed had clonal Igh gene rearrangements.

Given that the human disease most often associated with CALM-AF10 fusions is pre–T-LBL, we were somewhat surprised that the majority of the mice in this series had myeloid leukemia (although a minority of these leukemias had T-cell features such as CD3 staining and/or clonal Tcrb or Tcrd gene rearrangements). Because previously published models of pre–T-LBL have often shown evidence of perturbed thymocyte differentiation before the onset of pre–T-LBL, we studied thymocyte subsets in eight CALM-AF10 transgenic and eight WT control littermates. Supplementary Fig. S4A shows that four of eight transgenic mice had significantly decreased numbers of double-positive CD4⁺/CD8⁺ (hereafter DP) cells (1–36%), coinciding with dramatically increased numbers of more immature CD4⁻/CD8⁻ (hereafter DN) cells (27–90%). Consistent with this finding, a mouse with >80% DN cells (mouse no. 9045) had Tcrb alleles that were exclusively in the germ-line configuration, in contrast to control mice with normal (<10%) numbers of DN cells and transgenic mice with less marked (<30%) increases in DN cells, which showed polyclonal Tcrb gene rearrangements (Supplementary Fig. S4B). As shown in Supplementary Fig. S4, the degree of impaired thymocyte differentiation was quite variable, indicating that the penetrance of this phenotype was incomplete.

Because the mice developed primarily myeloid leukemias, we assayed the proportion of Mac1⁺/Gr1⁺ cells in the bone marrow to determine if there was an expansion of this population in clinically healthy mice. Although there were increased numbers of Mac1⁺/Gr1⁺ cells in the CALM-AF10 transgenic mice (68.7 ± 12.4) compared with WT littermates (54.2 ± 6.8), this difference did not reach statistical significance (Supplementary Fig. S5). We evaluated bone marrow from clinically healthy mice to determine if there was expansion of LSK cells in bone marrow. Although there was a trend toward decreased LSK cells in CALM-AF10 mice (0.03 ± 0.007%) compared with WT controls (0.08 ± 0.003%), this difference was not statistically significant.

When compared with pre–T-LBL patients without CALM-AF10 fusions, pre–T-LBL patients with CALM-AF10 fusions show up-regulation of HOXA cluster genes, including HOXA5, HOXA9, and HOXA10 (18). In addition, the CALM-AF10 patients showed up-regulation of BMI1, which is located within 500 kb of AF10 on chromosome 10. We assayed Hoxa5, Hoxa7, Hoxa9, Hoxa10, Hoxa11, Hoxa13, Hoxb4, Hoxd13, and Meis1 expression by real-time RT-PCR in leukemias from CALM-AF10 mice to determine if the up-regulation of these genes was seen in this mouse model as well as in human leukemias. As shown in Fig. 5, Hoxa5, Hoxa7, Hoxa9, Hoxa10, and Meis1 were all up-regulated in hematopoietic tissues (bone marrow, spleen, and thymus) from clinically healthy CALM-AF10 mice. Myeloid leukemias from CALM-AF10 mice also showed up-regulation of Hoxa5, Hoxa7, Hoxa9, Hoxa10, and Meis1, indicating that up-regulation of these genes occurs in myeloid as well as T-cell tumors associated with CALM-AF10 expression. Hoxb4 was modestly decreased in clinically healthy transgenic bone marrow as well as in one of four leukemia samples. Expression of Hoxa11, Hoxa13, and Hoxd13 was undetectable in all tissues and tumor samples after 35 cycles of amplification, in contrast to the threshold of detection for the 18S ribosomal control at 18 cycles of amplification. In contrast to the up-regulation of Hoxa cluster genes described above, conventional RT-PCR showed that Bmi1 was highly expressed in hematopoietic tissues and leukemias from both transgenic and nontransgenic mice, but not up-regulated (data not shown).

To develop a mouse model for CALM-AF10 leukemia, we generated transgenic mice that expressed the CALM-AF10 fusion gene under control of vav regulatory elements. These mice developed acute leukemia, after an extended latent period (median, 12 months), with an incomplete penetrance (40–50%). In addition, a substantial number of mice were found dead in their cages and were unable to be characterized as leukemic or nonleukemic. If all of these mice were leukemic, then the penetrance of the disease would increase to 70% to 83%, depending on the founder line, over the 18-month study period.

The acute leukemias that developed in CALM-AF10 transgenic mice were predominantly myeloid leukemias. Unexpectedly, half of these myeloid leukemias stained positive for B220. Although B220 is often regarded as a pan-B-cell marker, B220 expression has also been detected on natural killer, dendritic, and myeloid progenitors (19, 20). The Mac1⁺/B220⁺ leukemic cells were negative for more specific B-cell markers, such as CD19 and IgM, but were positive for CD24 and heterogeneous for CD117 expression. We studied Igh gene rearrangements as an additional marker of B-cell differentiation; more than half of the mice analyzed had clonal Igh rearrangements. Interestingly, the B220 staining did not correlate well with clonal Igh rearrangements; for instance, mouse no. 7160 was negative for B220 and had clonal Igh gene rearrangements, and mouse no. 9014 was positive for B220 and did not have clonal Igh gene rearrangements. In addition, two mice had a myelomonocytic leukemia, one of which stained weakly positive for CD3 and had both Tcrb and Tcrd rearrangements as well as an Igh rearrangement, and the other (mouse no. 7160) had a myeloid leukemia with both Tcrb and Igh rearrangements.

Although CALM-AF10 fusions have been identified in patients with myeloid, monocytic, and megakaryocytic leukemias (21–25), it was somewhat surprising that the mice in this study developed myeloid leukemia because CALM-AF10 leukemia in humans is primarily γ/δ pre–T-LBL. The observation that more than half (8 of 15) of the leukemias assayed had clonal Tcr or Igh gene rearrangements and more than half (9 of 17) stained positively for B220 suggests that the transformed cell, at least in some cases, was derived from a progenitor capable of lymphoid as well as myeloid differentiation. This finding is consistent with reports that clonal TCRD and/or IGH gene rearrangements are often detected in patients with CALM-AF10 fusion and myeloid leukemia (4, 26). Alternatively, the difference in leukemia phenotype may be due to use of the pan-hematopoietic vav regulatory elements to direct CALM-AF10 expression. In humans, it is possible that CALM-AF10 translocations occur only rarely in myeloid progenitors and more commonly in lymphoid progenitors. However, in CALM-AF10 mice, the vav transgene cassette directs expression to all hematopoietic cells (11, 27), including those in the myeloid compartment, allowing the CALM-AF10 fusion an opportunity to exert its oncogenic potential in myeloid cells. Finally, it is possible that a cis effect present in CALM-AF10 pre–T-LBL patients is missing in these transgenic mice, where the CALM-AF10 construct has integrated randomly. Support for this hypothesis comes from the observation that although patients with CALM-AF10 translocations have up-regulated BMI1, presumably through a cis effect because the AF10 gene is located near BMI1 on chromosome 10, the CALM-AF10 mice did not show up-regulated Bmi1. In any case, the results reported here show that expression of a CALM-AF10 fusion gene in myeloid cells predisposes these cells to leukemic transformation.

We studied healthy CALM-AF10 mice to determine if we could identify signs of impending leukemic transformation; however, CBCs from CALM-AF10 mice were not significantly different than those of their WT littermates (Supplementary Table S1). We detected abnormalities of thymocyte differentiation: half (4 of 8) of the CALM-AF10 mice had reduced numbers of DP thymocytes and increased numbers of more immature DN thymocytes. This impaired differentiation was reinforced by the observation that the mouse with the most dramatic increase in DN cells had no Tcrb rearrangements. Because most of the CALM-AF10 mice developed myeloid leukemia, we reasoned that clinically healthy mice might have an increase in the proportion of Mac1⁺/Gr1⁺ cells in the bone marrow. However, although there was an increased proportion of Mac1⁺/Gr1⁺ cells in the bone marrow of the CALM-AF10 mice compared with WT controls, this difference did not reach statistical significance.

Several lines of evidence developed over the past several years have shown the importance of HOX genes during leukemic transformation. First, multiple HOX genes have been identified as fusion partners with NUP98 in chromosomal translocations associated with acute leukemia (28). Second, gene expression profile studies have shown that up-regulation of several HOX genes, especially HOXA7 and HOXA9, occurs in both acute lymphoid and myeloid leukemia (29–31). Third, Hox genes, again including Hoxa7 and Hoxa9, have been found to be up-regulated by retroviral insertion in mice with retroviral-induced myeloid leukemia (32). Finally, a number of investigators have reported that overexpression of HOX genes or NUP98-HOX fusion genes leads to acute myeloid malignancies in mice (31, 33–36). Of note, recent reports showed overexpression of HOXA5, HOXA9, and HOXA10 in pre–T-LBL patients who had CALM-AF10 translocations (18, 29, 37). We show here that CALM-AF10 expression leads to up-regulation of Hoxa5, Hoxa9, and Hoxa10 in hematopoietic tissues from clinically healthy and leukemic CALM-AF10 transgenic mice. In contrast to previous studies that used retroviral transduction and transplantation to overexpress HOXA9 or HOXA10 in mouse bone marrow, we did not detect evidence for myeloproliferation in CALM-AF10 transgenic mice (38–42). This difference may be due to the relative levels of Hox gene expression, to the activation of multiple Hox genes by the CALM-AF10 transgene, or to the technique used to express CALM-AF10 in hematopoietic cells (retroviral transduction and transplantation versus transgenesis). Given the documented leukemogenicity caused by up-regulation of HOX genes, particularly those of the abd-b group (paralogues 9–13; ref. 31), it seems likely that the CALM-AF10 transgene exerts its leukemic effect, at least in part, through HOX gene activation.

While this article was under review, Deshpande et al. (26) reported the use of retroviral transduction and transplantation of primary bone marrow cells to show that expression of CALM-AF10 in mouse bone marrow cells led to myeloid leukemia. They noted that the myeloid leukemia cells were invariably B220⁺, with clonal Igh gene rearrangements, whereas in our study, only half of the myeloid leukemias were B220⁺ and only half had clonal Igh gene rearrangements. Similar to our findings, they found that the Mac1⁺/B220⁺ cells lacked specific B-cell markers such as CD19 and IgM. In contrast, the Mac1⁺/B220⁺ cells in their study lacked CD117 expression, whereas a variable (as high as 70%) percentage of Mac1⁺/B220⁺ leukemic cells from CALM-AF10 transgenic mice expressed CD117. Intriguingly, they noted that transplantation of Mac1⁻/B220⁺ leukemic cells led to myeloid leukemia in recipient mice, and that the leukemic stem cell was likely to be a Mac1⁻/B220⁺ cell. In the current study, we noted two distinct populations of Mac1⁻/B220⁺ cells in some of the CALM-AF10 leukemic mice (see Fig. 3). The bright Mac1⁻/B220⁺ population (Fig. 3A,, gate R4) likely reflected contaminating normal B-cells because these cells were positive for CD19 and IgM and were only seen in leukemic spleen but not leukemic bone marrow, whereas the dim Mac1⁻/B220⁺ cells (Fig. 3A , gate R5) were negative for CD19 and IgM and may represent part of the leukemic clone, similar to the B220⁺ cells transplanted by Deshpande et al. (26). Given the absence of other B lineage markers on the dim B220 population, along with reports of B220 expression on progenitor cells (20, 21), it is important to consider the possibility that the dim B220 expression detected in these leukemic samples does not indicate B-lineage commitment.

In this report, we have shown that CALM-AF10 expression is strongly leukemogenic. An emerging paradigm in leukemia biology predicts that most, if not all, leukemic cells must undergo at least two collaborative events to produce a fully transformed cell. One of these events leads to impaired differentiation and the second event results in increased proliferation and/or decreased apoptosis. The long latency period and incomplete penetrance of the leukemic phenotype support the hypothesis that at least one additional event is required to fully transform these cells. The impaired differentiation that we observed in the thymus, along with previous reports that HOX gene overexpression is associated with impaired blood cell differentiation (31), is consistent with a model in which overexpression of the CALM-AF10 gene impairs differentiation of hematopoietic cells. It seems likely that secondary, as yet undefined events that lead to increased proliferation and/or decreased apoptosis collaborate with the impaired differentiation caused by CALM-AF10 expression and result in completely transformed, leukemic cells.

Grant support: Intramural Research Program of the NIH, National Cancer Institute.

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 Michael Kuehl (National Cancer Institute, Bethesda, Maryland) for the gift of the murine Igh probe, Jerry Adams (Walter and Eliza Hall Institute, Melbourne, Australia) for the gift of the vav plasmid, and Lionel Feigenbaum (National Cancer Institute, Bethesda, Maryland) for generation of the transgenic mice, and Siba K. Samal, R. Mark Simpson, YingWei Lin, Chris Slape, and Helge Hartung for fruitful discussions and insight.