Hoxa9 collaborates with E2A-PBX1 in mouse B cell leukemia in association with Flt3 activation and decrease of B cell gene expression



Background: The fusion protein E2A-PBX1 induces pediatric B cell leukemia in human. Previously, we reported oncogenic interactions between homeobox (Hox) genes and E2A-PBX1 in murine T cell leukemia. A proviral insertional mutagenesis screen with our E2A-PBX1 B cell leukemia mouse model identified Hoxa genes as potential collaborators to E2A-PBX1. Here we studied whether Hoxa9 could enhance E2A-PBX1 leukemogenesis. Results: We show that Hoxa9 confers a proliferative advantage to E2A-PBX1 B cells. Transplantation experiments with E2A-PBX1 transgenic B cells overexpressing Hoxa9 isolated from bone marrow chimeras showed that Hoxa9 accelerates the generation of E2A-PBX1 B cell leukemia, but Hoxa9 is unable to transform B cells alone. Quantitative-reverse transcriptase polymerase chain reaction analysis demonstrated a strong repression of B cell specific genes in these E2A-PBX1/Hoxa9 leukemias in addition to Flt3 activation, indicating inhibition of B cell differentiation in combination with enhanced proliferation. Overexpression of Hoxa9 in established E2A-PBX1 mouse leukemic B cells resulted in a growth advantage in vitro, which was also characterized by an enhanced expression of Flt3. Conclusions: we show for the first time that Hoxa9 collaborates with E2A-PBX1 in the oncogenic transformation of B cells in a mouse model that involves Flt3 signaling, which is potentially relevant to human disease. Developmental Dynamics 243:145–158, 2014. © 2013 Wiley Periodicals, Inc.


The fusion protein E2A-PBX1 is generated through the translocation between chromosome 1 and 19 and is prevalent in 25% of pre-B cell leukemias in young children (Kamps et al., 1990; Nourse et al., 1990). E2A-PBX1 is a potent oncogene with transcriptional activating properties, which can transform a range of cell types, including fibroblasts, myeloid and lymphoid cells (Kamps et al., 1991; Kamps and Baltimore, 1993; Sykes and Kamps, 2004). Structure function analysis revealed critical elements of both genes for transformation of cells by the E2A-PBX1 fusion protein (Monica et al., 1994; Chang et al., 1997). The strong transcriptional activator properties of E2A-PBX1 are provided by the activation domains (AD) of E2A (van Dijk et al., 1993). Oncogenesis by E2A-PBX1 is dependent on AD1, which has been shown to interact directly with the KIX domain of transcriptional co-regulator CBP/p300 (Bayly et al., 2004). Disruption of the CBP binding to E2A-PBX1 by substitution of a single leucine residue in the LXXLL motif of AD1 impaired E2A-PBX1 induced immortalization of primary hematopoietic cells in vitro and development of myeloproliferative disease in bone marrow (BM) recipients (Bayly et al., 2004, 2006).

The E2A gene codes for basic helix-loop-helix proteins E12 and E47, which are essential for lymphoid development, inducing the expression of lymphoid specific genes such as recombinase activating genes, early B cell factor1 (Ebf1) and surrogate light chain λ5 (Bain et al., 1994; Quong et al., 2002). Activation of these genes results in the immunoglobulin rearrangement of B cells at the pre-B cell stage accompanied with the inhibition of proliferation (Quong et al., 2002; Mandal et al., 2009). Mice targeted at the E2A locus exhibit a complete block in B cell development beyond the pro-B cell stage and develop T cell leukemias indicating anti-proliferative and tumor suppressor properties of E2A (Bain et al., 1997). Thus heterozygosity of the E2A gene in the human disease may contribute to the development of leukemia.

The PBX1 fusion partner is a member of the TALE homeodomain transcription factors. These proteins can form complexes with pentapeptide containing homeobox proteins (paralog 1-10), increasing the DNA binding specificity of HOX proteins (Chang et al., 1995, 1996; Lu and Kamps, 1997). The transcriptional repressor/activation function of HOX/PBX1 heterodimers is determined by cell signaling events (Saleh et al., 2000). However, the fusion of the E2A activation domains to the C-terminal of the transcription factor PBX1 converts PBX1 to a constitutive transcriptional activator (van Dijk et al., 1993; Lu et al., 1994). Indeed, it has been demonstrated that E2A-PBX1 can bind to sequences recognized by HOX/PBX1 complexes (Chang et al., 1995; Lu et al., 1995; Phelan et al., 1995).

Domain requirements of E2A-PBX1 for oncogenic transformation appear to be cell type specific. Transformation of fibroblasts requires both AD1 and AD2 domains plus the Hox co-operating motif (HCM) of PBX1 (Chang et al., 1997), whereas transformation of myeloid cells requires, in addition, the homeodomain of PBX1 (Kamps et al., 1996). Requirement of the homeodomain for transactivation in myeloid transformation (Monica et al., 1994), suggests that deregulation of HOX/PBX1 elements or target genes plays a critical role in E2A-PBX1 induced myeloid leukemia. Moreover, retroviral overexpression of single Hox genes (e.g., Hoxa9, Hoxa10, Hoxb3, Hoxb6, and Hoxb8) is sufficient to induce leukemia in BM (Perkins and Cory, 1993; Thorsteinsdottir et al., 1997, 2001; Sauvageau et al., 1997; Kroon et al., 1998; Fischbach et al., 2005). It has recently been demonstrated that all Hoxa genes except Hoxa2 and Hoxa5 have oncogenic potential and could transform BM cells in culture (Bach et al., 2010). Co-overexpression of Hoxa9 and E2A-PBX1 results in a shorter latency to myeloid leukemia onset compared with Hoxa9 alone (Thorsteinsdottir et al., 1999), but interestingly Hoxa9 did not collaborate with E2A-PBX1 in mouse T cell leukemia (Bijl et al., 2008).

So far, it is unclear whether oncogenic interactions between Hox genes and E2A-PBX1 occur in development of clinically relevant B cell leukemia. We have previously generated a lymphoid specific transgenic mouse model for E2A-PBX1 that generates pre B-acute lymphoblastic leukemia (B-ALL), phenotypically similar to human disease, in the absence of T cell maturation (Bijl et al., 2005). Evidence for a potential role of Hox genes in E2A-PBX1 induced B cell leukemia was obtained from a proviral insertional mutagenesis screen using this model. A hot spot for viral integrations in the Hoxa locus in association with the presence of E2A-PBX1 that resulted in an overall activation of the Hoxa cluster genes in leukemic cells strongly suggest collaboration of Hoxa genes with E2A-PBX1 in B cell leukemogenesis.

To further investigate the importance of Hox collaboration with E2A-PBX1 in the development of B-ALL, Hoxa9 was overexpressed in nonmalignant and malignant B cells derived from E2A-PBX1/CD3ε−/− transgenic mice. In this study we provide evidence that Hoxa9 collaborates with E2A-PBX1 in oncogenic transformation of B cells, which might be mediated through efficient repression of B cell specific genes and activation of Flt3, a direct target of Hoxa9 (Wang et al., 2006). Moreover, Hoxa9 also promoted the growth kinetics and leukemia regeneration potential of primary E2A-PBX1 induced B cell leukemias that involved changes in expression of genes implicated in hematopoietic differentiation, in particular Flt3, Pdgfδ, and Lmo1. Here we show for the first time the oncogenic transformation of B cells through interactions between Hox genes and E2A-PBX1 involving deregulation of developmental programs.


Hoxa Gene Expression in Malignant and Nonmalignant E2A-PBX1 B Cells

The expression of all eleven HOXA genes was measured in four human B cell leukemia samples harboring the fusion E2A-PBX1 and two without E2A-PBX1 by quantitative-reverse transcriptase polymerase chain reaction (Q-RT-PCR). In addition, RNASeq data from 23 pre-B ALLs without known major translocations were available. All E2A-PBX1 samples expressed the majority of HOXA genes, but the expression levels for the individual genes were very variable (Fig. 1A). The expression of HOXA genes in non-E2A-PBX1 B-ALL was in general negligible to low/medium (Fig. 1A, C, D). Contrarily to the non-E2A-PBX1 B-ALLs, at least one HOXA gene was expressed at high levels in individual E2A-PBX1 leukemias, and notably HOXA7 expression was constantly high across the E2A-PBX1 samples. Intriguingly, one E2A-PBX1 leukemia sample showed high expression levels of almost the entire HOXA cluster, resulting in one log higher accumulated HOXA gene expression (Fig. 1B).

Figure 1.

Hoxa gene expression in human and mouse E2A-PBX1 leukemias. A: Heatmap of HOXA genes expression in human pediatric E2A-PBX1+ (n = 4) and E2A-PBX1- (n = 2) B-ALLs. B: Cumulated HOXA genes expression in human B-ALLs. Standard curve generated copy numbers for each group are calculated from the average CT values normalized for GAPDH (CT = 23) in each group according to the following formula 2(38-CT). C,D: Average HOXA gene expression in human pre-B ALLs positive for E2A-PBX1 (n = 4) obtained by Q-RT-PCR (C) and without known major translocations (n = 23) obtained by RNASeq (D). The CT values (Q-RT-PCR), corrected for GAPDH expression at CT = 20, were related to the RPKM values (RNASeq) according to the following formula logRPKM = −0.17*(ΔCT) + 2.4. This formula has been empirically determined based on comparison of Q-RT-PCR data and RPKM data from the same sample (Sauvageau's Lab, personal communication). The threshold of HOX gene expression is considered around CT = 35, which corresponds to a RPKM value of 0.7 (indicated by dotted line). Note that E2A-PBX1 pre-B ALLs express all HOXA genes. The only pre-B ALL that showed expression of several HOX genes had an amplification of RUNX1. RPKM = Reads per kilobase of transcript per million mapped reads. B-ALL = B cell acute lymphoblastic leukemia.

The Hoxa gene expression levels were also measured in B cell leukemias derived from spleen of E2A-PBX1/CD3ε−/− transgenic mice and compared with the expression in Moloney Murine Leukemia Virus (MMLV) induced B cell leukemias. Hoxa gene expression was low to medium/high in nine E2A-PBX1 positive leukemias and severely deregulated in two (∼ one log higher) (Fig. 2A). The expression levels for Hoxa genes were also low to medium/high in MMLV induced B cell leukemias, but extreme high expressions were not observed. Expression of Hoxa genes in E2A-PBX1 leukemic cells tended to be higher than in B cells from healthy preleukemic E2A-PBX1 transgenic mice (Fig. 2B). Of interest, the levels of Hoxa gene expression were also occasionally very high in B cells sorted from BM of preleukemic E2A-PBX1 transgenic mice (Fig. 2C). Overall the expression of Hoxa genes is clearly present in E2A-PBX1 transgenic malignant B cells and reflects those in immature B cells found in the BM. The aberrant high Hoxa gene expression in a subset of E2A-PBX1 leukemias suggests that the Hox pathway might contribute to E2A-PBX1 leukemogenesis.

Figure 2.

A,B: Heatmaps of Hoxa genes expression in leukemias from (A) E2A-PBX1/CD3ε−/− transgenic mice (n = 11) and MMLV induced B-ALLs (n = 8) and of (B) Nonmalignant B220+ B cells purified from spleen of healthy 3–5 months old E2A-PBX1/CD3ε−/− transgenic and wild-type mice. C: Heat map of Hoxa gene expression in nonmalignant B cells sorted from bone marrow of 3- to 5-month-old healthy E2A-PBX1/CD3ε−/− transgenic (n = 4) and wild-type mice (n = 4). Note the high expression of Hoxa genes in one of the E2A-PBX1 positive samples (lane 4). B-ALL, B-acute lymphoblastic leukemia; MMLV, Moloney Murine Leukemia Virus. nd, not determined.

Ex Vivo Cultures of Hoxa9 Transduced E2A-PBX1 B Cells

To directly test the effect of high Hox gene expression on the growth response to IL-7 of E2A-PBX1 B cells, B220+ cells were sorted from BM of three to five month old healthy / preleukemic transgenic mice. Q-RT-PCR analyses verified that the basal Hoxa gene expression levels in these B cells were low to moderate (data not shown). Isolated cells were transfected with Hoxa9 or green fluorescent protein (GFP) retroviral vectors and subsequently grown in B cell supporting medium (Fig. 3A). In each of three independent experiments Hoxa9 transduced E2A-PBX1 B cells grew faster than control, reaching expansions up to 800-fold over the initial population within 20 days (Fig. 3B and data not shown). Maximum expansion in these cultures, observed at 11 days, was only eight-fold. Doubling times, calculated based on the exponential growth of the cultures, were consistently one to five days shorter for Hoxa9 transduced E2A-PBX1 B cell cultures than its corresponding controls (Fig. 3C; P = 0.03). The expansion of the E2A-PBX1/Hoxa9 B cell cultures was supported by an increasing number of clonogenic progenitors, while in control cultures progenitors numbers were significantly lower after day 9 (Fig. 3D). Immunophenotyping by flow cytometry showed that the E2A-PBX1 cells from both control and Hoxa9 transduced cultures were B220+/CD43+/IgM-/BP1+ (Fig. 3E). To exclude whether Hoxa9 alone could confer such a growth expansion within the same magnitude, similar cultures were initiated with CD3ε−/− control B cells. Although Hoxa9 cultures grew faster than control, only a maximum of four-fold expansion was observed (at day 11) followed by a decline (data not shown). Thus Hoxa9 and E2A-PBX1 together provide B cells a strong proliferation capacity in vitro, which could be indicative for their collaboration in B cell leukemia induction.

Figure 3.

Analysis of E2A-PBX1/CD3ε−/− B cell cultures overexpressing Hoxa9 or control. A: Schematic of the experimental strategy and representation of Hoxa9-GFP and control-GFP retroviral vectors. Restriction sites relevant for this study are indicated. R, EcoRI; N, NcoI. B: Representative growth curves of E2A-PBX1/CD3ε−/− pro-B cells transduced with Hoxa9 or GFP (n = 3). C: Bar graph presenting the doubling time for Hoxa9 and control pro-B cell cultures (P = 0.03). D: Total IL-7 responsive B cell progenitors in the pro-B cell cultures transduced with Hoxa9 or GFP measured by colony forming cell at indicated time points (n = 2). *P ≤ 0.05 1-tailed Student's t-test. E: FACS profiles for B cells surface markers involved in staging the B cell differentiation confirming the pro-B cell phenotype (B220+/CD43+/IgM-/BP1+) of Hoxa9 and control E2A-PBX1/CD3ε−/− cultured cells. IgM, Immunoglobulin M; BP1, B cell Protein1. B-CFC, B cell colony forming cell.

Evaluation of Interaction Between E2A-PBX1 and Hoxa9 in B Cell Leukemogenesis

E2A-PBX1 transgenic BM chimeras were generated overexpressing either Hoxa9 or, GFP. Retroviral overexpression of Hoxa9 in total BM cells is known to give rise to a myeloid leukemia in transplantation recipient mice within five months (Kroon et al., 1998). Therefore, all chimeras were killed after 1 month and GFP+ donor derived E2A-PBX1 B cells were purified from the BM and transferred into irradiated recipients together with supporting BM helper cells (Fig. 4A). In a first experiment, mice that received E2A-PBX1/Hoxa9 B cells (n = 2) developed leukemia within three months (Table 1), while recipients of E2A-PBX1/GFP B cells (n = 3) did not show signs of leukemia after four months at which time no GFP+ cells could be detected. FACS analysis showed that E2A-PBX1/Hoxa9 leukemic cells from both mice expressed CD43 and variable percentages of B220 and Mac-1 (Fig. 4B and data not shown), indicating a potential bi-phenotypic origin of these leukemias. Transfer of leukemic cells into mice resulted in generation of secondary leukemias with reduced number of B220+ cells (Fig. 4B). As the change in immunophenotype of the leukemias could result from the presence of minor clones that became more prominent, leukemias were analyzed for their clonal composition. Southern Blot analysis showed that leukemias were monoclonal as the same clone was detected in all five transplantation recipients (Fig. 4C). Thus the loss of B220 expression in secondary leukemias was not the result of the selective advantage of a B220 negative clone. To prevent transformation of a primitive B220+/Mac-1+ B cell by Hoxa9, a second experiment was conducted using sorted B cells that were negative for Mac-1. Recipients of both Hoxa9 (n = 7) and control (n = 5) E2A-PBX1 B cells developed leukemia (Table 1). However, in the context of Hoxa9 the average latency time to development of disease was significantly reduced compared with control (survival 35.7 ± 17.6 for E2A-PBX1/Hoxa9 vs. 48.8 ± 5.2 days for E2A-PBX1/GFP; P = 0.04; Table 1; Fig. 4D). Mice transplanted with Hoxa9 transduced wild-type B cells did not succumb of disease within the timeframe observed. FACS analysis confirmed a B cell phenotype for all the leukemias as demonstrated by high expression of CD19. However, the pan B cell marker B220 was consistently much lower expressed in the majority of both E2A-PBX1/Hoxa9 and E2A-PBX1/GFP leukemias (Fig. 4E). Interestingly, several leukemias of both groups exhibited Mac-1 expression on a small percentage of the cells (Table 2). To test whether the leukemias could be re-initiated with maintenance of the phenotype, 1-2 × 106 cells of leukemias induced by E2A-PBX1/Hoxa9 (n = 3) or E2A-PBX1/GFP (n = 2) were injected into recipients (n = 2 for each leukemia). All mice succumbed to leukemia between two and four weeks posttransplantation (data not shown). Several E2A-PBX1/Hoxa9 and E2A-PBX1/GFP re-initiated leukemias cells showed lower B220 expression compared with primary leukemias. In addition, expression of the B cell marker CD19 was also reduced in most cases (Table 2). Recently it has been reported that Hoxa9 is inversely expressed during B cell differentiation with Ebf1 (Gwin et al., 2010). Hoxa9 appeared to be required for the expansion of pro-B cell populations, mediated through direct activation of Flt3. Therefore, we sought to analyze whether Hoxa9 expression in the context of E2A-PBX1 also induced Flt3 expression and could affect B cell differentiation genes Ebf1 and Pax5. Q-RT-PCR analysis showed that Flt3 was 16-fold and 4-fold higher expressed in E2A-PBX1/Hoxa9 leukemias compared with E2A-PBX1 (P = 0.003) or MMLV induced B cell leukemias (P = 0.03; Fig. 5), respectively. A large difference in Flt3 expression was also observed with Hoxa9 induced myeloid leukemia. In addition, expression of B cell differentiation genes Pax5 and Ebf1 was reduced in E2A-PBX1/Hoxa9 leukemias reaching significance compared with MMLV (P = 0.022 and 0.015, respectively) and close to significance compared with E2A-PBX1 induced leukemias (P = 0.055 for Pax5 and P = 0.074 for Ebf1). Expression analysis of nonmalignant B cells sorted from Hoxa9 BM chimeras showed a two-fold increase in Flt3 expression compared with GFP B cells, but no difference in Ebf1 and Pax5 expression was observed. Thus these experiments show that Hoxa9 can collaborate with E2A-PBX1 in B cell leukemogenesis, but does not transform B cell alone, and that deregulation of B cell specific genes might be important in this oncogenic process.

Table 1. Collaboration Between Hoxa9 and E2A-PBX1/CD3ϵ-/- in B Cell Leukemogenesis
GenotypeMSCV vectornCell populationCell DoseLeukemic mice (%)Survival (days)
  1. a

    Mice were sacrificed after 4 months.

  2. b

    P = 0.04 1-tailed Student's t-test; a one-tailed test was chosen based on the hypothesis that Hoxa9 accelerates E2A-PBX1 leukemia. na = not applicable.

Experiment 1
E2A-PBX1/CD3ɛ−/−Hoxa9-GFP2B220+3×10510076 ± 14.1
Experiment 2
E2A-PBX1/CD3ɛ−/−Hoxa9-GFP7B220+/Mac-1-3×10510035.7 ± 17.6a
E2A-PBX1/CD3ɛ−/−GFP5B220+/Mac-1-5×10510048.8 ± 5.2a
Table 2. Immunophenotype of Primary and Their Corresponding Secondary Leukemias Expressing Hoxa9 or GFP
 LeukemiaDonor cellsB220CD19CD43Mac-1
Secondary#82 10.376.720.317.7
 #85 25.469.734.427.9
Secondary#76 20.259.950.38.7
 #77 29.575.158.46.6
Secondary#86 37.714.310.91.4
Secondary#70 22.760.551.635.9
 #71 21.351.651.233.9
Figure 4.

In vivo analysis of Hoxa9 or control transduced E2A-PBX1/CD3ε−/− B cells. A: Schematic of the experimental strategy. B: FACS profile of primary leukemia #101 (Experiment 1) and two secondary leukemias regenerated from #101 showing surface markers CD43 and B220. Plots are gated on GFP+ cells. Note that expression of B220 is lower in the secondary leukemias. C: Clonal analysis of secondary recipients of E2A-PBX1/CD3ε−/− leukemia #101 by Southern blot on analysis on genomic DNA isolated from primary and secondary leukemias. Note that the same sets of fragments of equal intensity for all chimeras indicate their monoclonal origin. D: Kaplan-Meier survival curves of mice transplanted with E2A-PBX1/CD3ε−/− B cells transduced with Hoxa9 (n = 7) or GFP (n = 5) and CD3ε−/− B cells transduced with Hoxa9 (n = 2) or GFP (n = 9). Hoxa9 significantly accelerates mean latency to E2A-PBX1 leukemia, P = 0.04, 1-tailed Student's t-test, but curves are not significantly different Chi Square test P = 0.08. E: FACS analysis of E2A-PBX1/Hoxa9 and E2A-PBX1 induced B cell leukemias.

Figure 5.

Differential expressions of Flt3, Ebf1, and Pax5 in E2A-PBX1/Hoxa9 leukemias. Heat map for expressions of these genes in B cell leukemias induced by MMLV (n = 2), E2A-PBX1 (n = 4) or E2A-PBX1/Hoxa9 (n = 3), and in nonmalignant B cells transduced with Hoxa9 (n = 2) or control GFP (n = 2) was generated with DataAssist v3.0 (Applied Biosystems). Each row presents the expression of a different gene. Average cycle threshold (CT)-values for these genes are given below for each of these groups and for Hoxa9 induced AML as reference. High CT-values representing low expression levels are presented in green, while low CT-values indicting high expression levels are indicated in red. MMLV, Moloney Murine Leukemia Virus. AML = acute myeloid leukemia.

Hoxa9 Overexpression in E2A-PBX1 Primary Leukemias

In addition to its oncogenic interactions with E2A-PBX1 in leukemia induction, we questioned whether Hoxa9 could enhance the leukemic properties of established E2A-PBX1 leukemias. Primary E2A-PBX1 B cell leukemias generated from our transgenic mouse line were transduced with Hoxa9-GFP or control GFP retroviral vectors and grown on S17 feeder cells in the presence of lymphoid specific growth factors (Fig. 6A). Three of nine leukemias transduced with control vector and five out nine transduced with Hoxa9 were able to grow under these conditions (Table 3), indicating that Hoxa9 enhances the culturing of E2A-PBX1 leukemic B cells. Moreover, the expansion of E2A-PBX1 B cell leukemia cultures was significantly enhanced in the presence of Hoxa9 for two out of the three leukemias growing under both conditions (#225 and #282, Table 3), exemplified by an increasing ratio of Hoxa9 vs control cells for leukemia #225 (Fig. 6B), and a slightly but significantly reduced doubling time (30.7 ± 0.9 and 28.2 ± 0.5 hr for control and Hoxa9, respectively, P=0.04; Fig. 6C). This growth advantage of Hoxa9 transduced leukemic cells was supported by an increase in leukemic B cells with clonogenic properties, which are expected to maintain the culture as their normal B progenitor equivalents (Fig. 6D). Compared with B cell colonies induced by nonmalignant progenitors, leukemic colonies were more disorganized and morphologically aberrant (Fig. 6E). However, FACS analysis showed no differences in the common B cell surface markers in the presence of Hoxa9 (data not shown). These data show that overexpression of Hoxa9 in established E2A-PBX1 leukemic B cells could enhance growth capacity and proliferation kinetics of several primary E2A-PBX1 leukemias.

Table 3. Hematopoietic Characterization and In Vitro Growth Potential of Primary E2A-PBX1 B-ALL
MouseAge (days)WeightSpleen (g)Phenotype leukemic B cellsGrowth
  1. a

    At the moment of sacrifice mouse did not show signs of disease.

  2. nd = not determined.

Figure 6.

Analysis of E2A-PBX1/CD3ε−/− leukemic cells overexpressing Hoxa9 or control. A: Schematic of the experimental strategy. B: Fold difference in growth of E2A-PBX1/CD3ε−/− leukemic cells #225 transduced by Hoxa9 over control at given time points. C: Doubling times for E2A-PBX1 leukemia #225 overexpressing Hoxa9 or control (n = 2). D: Total of clonogenic B cell progenitors in cultures of E2A-PBX1/CD3ε−/− leukemia #225 overexpressing Hoxa9 or control measured by colony forming cell at indicated time points. E: Example of B cell colony initiated by E2A-PBX1 leukemic clonogenic B cell progenitors (left). As a reference a typical round shaped colony initiated by nonmalignant B cell progenitors is shown (right). *P ≤ 0.05 1-tailed Student's t-test. B-CFC = B cell colony forming cell.

Gene Expression Profiling of E2A-PBX1 Primary Leukemias Overexpressing Hoxa9

To analyze which genes could underpin the observed changes induced by Hoxa9, expression analysis was performed by Q-RT-PCR on Hoxa9 and GFP cultured E2A-PBX1 leukemia cells (#111 and #225) using TaqMan primer/probe sets for 110 genes including several genes important for lymphoid and myeloid differentiation (Supp. Table S1, which is available online). Significant increased expression (Log10 >0.3) was observed for twelve genes (#225) or 13 genes (#111) in the presence of Hoxa9 (Table 4). The expression of only two genes, Flt3 and Pdgfd, was significantly and consistently increased in both Hoxa9 transduced leukemias compared with their respective controls. Flt3 has been previously reported as a direct target of Hoxa9 and is known for its involvement in leukemia. In addition a set of fourteen genes was significantly down-regulated (Log10 < −0.3) in Hoxa9 leukemia #225, while only three genes in Hoxa9 leukemia #111 (Table 4). One of these genes, Lmo1 a Lim domain protein involved in transcriptional regulation, was common to both leukemias. Surprisingly, the expression of seven genes, including Pdgfb, Col1a1, Pparg, Timp2, Ctnna, Gas6, and Sox4 were differentially expressed in Hoxa9 leukemia cells #111 and #225. Of these genes Sox4 has been recently shown to be regulated by Hoxa9 (Huang et al., 2012). Up-regulated genes by Hoxa9 in leukemia #225 include several transcription factors, such as Runx3 and Cebpα, which are genes associated to the myeloid lineage. In Hoxa9 #111 leukemic cells the up-regulation of two collagen encoding genes (Col3a1 and Col1a1) and two genes related to cell adhesion (Timp2 and cadherin Ctnna) was striking. Noteworthy is that adhesion molecules have been shown to be targets of Hox genes and Col3a1 has been proposed to be a direct target gene for Hoxa9 in adipocytes. These data show that Flt3 is a consistent target of Hoxa9 in malignant B cells and that the expression of several genes associated to hematopoietic differentiation is changed in the presence of Hoxa9.

Table 4. Differential Expressed Genes in Hoxa9 Transduced Leukemias #225 and #111
Leukemic cells #225Leukemic cells #111
  1. Note that gene expression is considered to be up-regulated in context with Hoxa9 when Log10(RQ) values are > 0.3 and down-regulated when values are < −0.3.

Up-regulated genes
AgerDevelopment, inflammation0.83Col3a1Extracellular matrix1.84
Flt3Hematopoietic differentiation0.79Flt3Hematopoietic differentiation0.49
Id3Repression lymphocyte differentiation0.64PpargTranscriptional regulation1.41
PdgfdProliferation and differentiation0.51PdgfdProliferation and differentiation0.34
Runx3Myeloid differentiation0.51Gas6Cell growth and survival0.72
CebpaMyeloid differentiation0.50MmeCleavage0.64
Cdkn2aCell cycle regulation0.48pdgfbProliferation and differentiation0.59
Gata3T-cell differentiation0.44Timp2Cell adhesion0.48
Ebf1B-cell differentiation0.44Col1a1Extracellular matrix0.40
CybbNeutrophil function0.38Sox4B-cell differentiation0.38
BLKB-cell differentiation0.35Mta2Transcriptional regulation0.32
Cdkn1bCell cycle regulation0.35CtnnaCell adhesion0.31
   Cdk4Cell cycle regulation0.31
Down-regulated genes
Lmo1Transcriptional regulation−0.40Lmo1Transcription regulation−0.52
Fn1Cell adhesion−1.36rpl3aTranslation−0.32
PdgfbProliferation and differentiation−1.31   
Ssp1Differentiation/immune response−0.95   
Cola1a1Extracellular matrix−0.85   
PpargTranscription regulation−0.80   
Timp2Cell adhesion−0.74   
CtnnaCell adhesion−0.58   
Gas6Cell growth and survival−0.53   
Sox4B-cell differentiation−0.49   
Mdm2p53 degradation/cell survival−0.36   
ItgamMyeloid cell function/adhesion−0.32   


Aberrant Hox gene expression is associated with leukemias, most notably those that harbor fusion proteins involving MLL (Armstrong et al., 2002; Ferrando et al., 2003). Large expression screens have not yet associated altered HOX gene expression with pediatric pre-B cell leukemias positive for the t(1;19) translocation encoding for E2A-PBX1 (Yeoh et al., 2002). This is attributable to incomplete probe set coverage of the HOXOME on these arrays and thus does not exclude abnormal HOX gene expression levels in these leukemias. In this study we show that E2A-PBX1 induced B cell leukemias generated in mice using a transgenic model express most of the Hoxa genes, which are at significantly higher levels than B-ALLs induced by MMLV. The inconsistent aberrantly high levels of Hoxa gene expression seen in both mouse and human E2A-PBX1 leukemias indicate that HOXA genes might indeed play an important role in the development of this disease. It is not clear why these HOXA gene levels are not consistently high in all E2A-PBX1 leukemias. It is possible that HOX genes from other clusters are aberrantly expressed or non-HOX/PBX pathways might be more prominently contribute to E2A-PBX1 leukemia and do not require high HOX expression levels. A decrease in HOX gene expression is normally observed with the maturation of hematopoietic cells (Sauvageau et al., 1994; Lebert-Ghali et al., 2010) and thus maintenance of moderate HOX gene expression levels might be sufficient to block differentiation as observed for Hoxa9 overexpression at the pre-B cell stage (Thorsteinsdottir et al., 2002).

Our data also show an impressive expansion of E2A-PBX1 pro-B cells by Hoxa9 in vitro at a magnitude that was not observed for B cells overexpressing either E2A-PBX1 or Hoxa9 individually. The synergistic actions of E2A-PBX1 and Hoxa9 in proliferation suggest that they activate complementing pathways. Also this is the first time that Hoxa9 has been shown to give a proliferative advantage to primary pro-B cells. These data are consistent with earlier reports that Hoxa9 is normally expressed in pro-B cells and is required for the generation of B cell progenitors (Lawrence et al., 1997; Gwin et al., 2010).

Retroviral and transgenic mouse models have shown genetic interactions between E2A-PBX1 and Hoxa9 or Hoxb4 in myeloid and T cell leukemia development, respectively (Thorsteinsdottir et al., 1999; Bijl et al., 2008). In the present study we show for the first time collaboration between a Hox gene and E2A-PBX1 in the transformation of B cells. Despite the fact that Hoxa9 is a potent oncogene, inducing myeloid leukemia (Kroon et al., 1998; Thorsteinsdottir et al., 2001), we showed that Hoxa9 induced transformation of B cells is dependent on the presence of E2A-PBX1. This is in accordance with the absence of lymphoid leukemia development in lymphoid specific Hoxa9 transgenic mice (Thorsteinsdottir et al., 2002). Of interest, our data show that Hoxa9 can transform advanced hematopoietic progenitors in the presence of the right collaborating oncogene, in this case E2A-PBX1. This is in contrast to findings in the myeloid environment where potent oncogenic combination of Hoxa9 and Meis1 efficiently transformed cells from the stem cell fraction LKS, but did very poorly in transformation of c-kitlo granulocyte macrophage progenitors (Wang et al., 2010). This study also reported that the Hoxa9/Meis1 or MLL/AF9 transformation events are dependent on the presence of active β-catenin. Of interest, a member of the Wnt family, Wnt16b, has been previously identified as a putative target of E2A-PBX1 (McWhirter et al., 1999), which expression appeared essential for the survival of human cell lines with the rearrangement t(1;19) (Mazieres et al., 2005). Also it has been shown that pro-B cells are dependent on Wnt signaling (Reya et al., 2000, Staal and Clevers, 2005). It is, therefore, not surprising that our E2A-PBX1/Hoxa9 leukemias displayed predominantly a pro-B cells (B220+/CD43+/IgM-) phenotype. The presence of Mac-1 on a low percentage of leukemic cells, which increased in secondary leukemias of some E2A-PBX1/Hoxa9 or E2A-PBX1/GFP primary leukemias is likely the result of granulocyte-colony stimulating factor receptor expression, which is a target of E2A-PBX1 (de Lau et al., 1998).

The Q-RT-PCR results suggest that signaling through Flt3 may play a role in the acceleration of E2A-PBX1 leukemia onset by Hoxa9 as well as in the enhanced proliferative properties of Hoxa9 transduced established E2A-PBX1 induced leukemias. However, increased Flt3 expression in Hoxa9 transduced control B cells that fail to develop leukemia indicates that activation of this pathway alone is not sufficient to transform B cells. The strong down-regulation of Ebf1 and Pax5 expression suggests that repression of B cell differentiation in combination with enhanced proliferative signals provided by Flt3 together might mediate the oncogenic interactions between Hoxa9 and E2A-PBX1 in leukemogenesis. In the Flt3 promoter binding regions for Hoxa9 and Pbx/Meis1 have been determined and Hoxa9 has been shown to activate transcription of Flt3 (Wang et al., 2006; Gwin et al., 2010; Volpe et al., 2013). Although no evidence exist regarding promoter binding by Pbx or E2A-PBX1 it is not excluded that Hoxa9 and E2A-PBX1 both bind as a complex to the Flt3 promoter, resulting in a strong activation of transcription. We do not expect that Meis1, which is a collaborator oncogene to Hoxa9 in myeloid leukemia, plays a critical role in E2A-PBX1 B-ALL. Meis1 is expressed in E2A-PBX1 and MMLV mouse leukemias, but its expression is not significantly different in the context with Hoxa9 (Fig. 7). Also contrarily to Pbx1, Meis1 is normally associated with activation of transcription (Sagerstrom, 2004; Choe et al., 2009) like E2A-PBX1 and, therefore, its presence might be redundant.

Figure 7.

Expression of Meis1 in mouse B cell and myeloid leukemias obtained by Q-RT-PCR. Values are corrected for GAPDH CT = 19.

Finally, we show that the proliferation of primary E2A-PBX1 B cell leukemias was enhanced in the presence of Hoxa9. Our targeted expression array showed that Flt3 and Pdgfδ could contribute to these biological functions. In addition, genes associated with myeloid differentiation, such as Cebpα and Runx3, were higher expressed in the context of Hoxa9 in these leukemic cells, which concurs with the reported requirement for Hoxa9 in myeloid differentiation (Lawrence et al., 1997) and its ability to induce myeloid leukemia in mice (Kroon et al., 1998; Thorsteinsdottir et al., 2001). Furthermore, Hoxa9 induced up-regulation of Id3, a gene associated with repression of lymphoid differentiation, fits the current view that Hoxa9 needs to be down-regulated for B cell specification by Ebf1 and Pax5. The fact that lymphoid specific transgenic mice for Hoxa9 have a partial block at the pre-B cell stage corresponds with this view (Thorsteinsdottir et al., 2002). It remains puzzling why a set of seven genes, including Sox4, demonstrates differential expression in two primary E2A-PBX1 leukemias. It is known that HOX-PBX complexes might act both as transcriptional activators as well as repressors, which are dependent on recruitment of additional factors, and thus suggest that the molecular profile of these two leukemias might be different. It is of note that the major site (and potentially the site of origin) of leukemia #111 was located in the liver in contrast to leukemia #225 that had principal involvement in the spleen, which might explain the differences in expression of the extracellular matrix components and adhesion molecules. In line with such a context dependent transcriptional regulation is that Pim1, a gene regulated by Hoxa9 in myeloid cells is not differentially expressed in Hoxa9 overexpressing leukemic B cells.

In conclusion our data show oncogenic interactions between Hoxa9 and E2A-PBX1 in transformation of B cells that may be mediated through complementing pathways involving proliferation conferred by Flt3 and Pdgfδ signaling and inhibition of B cell developmental programs. The importance of such HOX associated pathways for maintenance of E2A-PBX1 B cell leukemias remains to be determined.



Trizol (Invitrogen Corporation, Carlsbad, CA), samples of human pediatric ALL harboring the t(1;19) translocation were obtained from the Biobank of the Sainte-Justine Hospital (Montreal). Two control ALL samples for Q-RT-PCR were obtained from the Banque de cellules de Leucemiques du Quebec (BCLQ; www.bclq.org), while RNASeq data were obtained from 23 pre-B ALL samples (FAB L1). No translocations were detected in control ALLs. A standard Ficoll-Pacque gradient was performed on human samples to isolate mononuclear cells. All human samples contained over 70% of blasts. C57Bl6 inbred wild-type mice were purchased from Jackson Laboratories (Bar Harbor, ME). Transgenic mice for the E2APBX1 fusion protein were kept on a CD3ε−/− background as previously described (Bijl et al., 2005). Mice were bred and maintained in a specific pathogen free animal facility of the HMR Research Center. For the induction of B cell leukemias, 1-day-old newborn CD3ε−/− mutant mice were intraperitoneally injected with 10 infectious units of MMLV (van Lohuizen et al., 1991; Bijl et al., 2005). All animal protocols were approved by the Animal Care Committee of the HMR Research Center.

Quantitative RT-PCR

Total RNA was isolated by Trizol reagent (Invitrogen) in combination with RNeasy® clean up columns (Qiagen, Toronto, ON), DNase-I-treated and cDNA was prepared using MMLV-Reverse Transcriptase and random primers (all Invitrogen) according to the manufacturer's protocols. Q-PCR was carried out using SYBR Green chemistry (Applied Biosystems, Toronto, Canada) on an ABI 7500 thermal cycler (Applied Biosystems). Oligonucleotides for all 11 murine and human Hoxa genes were used according to previously described and validated sequences (Thompson et al., 2003; Dickson et al., 2009), and were tested to be used with SYBR Green (Lebert-Ghali et al., 2010). Validated TaqMan GEx assays for the other genes of interest were obtained by Applied Biosystems. Only replicate cycle threshold (CT)-values within 0.5 CT are accepted for Hox genes and 0.2 CT for the endogenous control GAPDH. Copy numbers for Hoxa genes are calculated from the average CT values in each group according to the following formula 2(38-CT). Values less than 10 (∼CT = 35) are considered not expressed.

Flow Cytometry and Cell Sorting

Phenotypic characterizations of B cell cultures were performed using the following conjugated antibodies: CD45R/B220-APCCy7, CD43-APC, IgM-biotin, CD11b (Mac-1)-PE (BioLegend, San Diego, CA). Biotinylated antibodies were detected with PerCP5.5 conjugated streptavidin (BioLegend). Mortality levels were determined using Dapi (Invitrogen). FACS analyses were performed on a FACS LSRII with FACSDiva software (BD Bioscience, Mississauga, ON). Data were further analyzed using FlowJo software (Tree Star Inc., Ashland, OR). All cell sortings have been performed on a FACS Aria II with FACSDiva software (BD Bioscience).

B Cell Cultures and Retroviral Infections

E2A-PBX1 preleukemic and leukemic B cells were prestimulated overnight in the presence of IL-7 or a cocktail of IL-7, Flt3 (Orf Genetics, Reykjavik, Iceland) and Steel factor (all at 10 ng/ml), respectively, followed by retroviral gene transfer during three day co-cultivation on packaging cell line (GP +E-86) (Markowitz et al., 1990) engineered to stably express MSCV-Hoxa9-GFP or control GFP retroviruses. Preleukemic cells were then cultured in OptiMem (Invitrogen) media supplemented with IL-7 (10 ng/ml; Invitrogen), 10% B cell tested Fetal Bovine Serum (Stem Cell Technologies, Vancouver, CA), 5 × 10−5 M 2-Mercaptoethanol (Mallinckrodt Baker Inc., Phillipsburg, NJ), 1× Penicillin-Streptomycin and 50 μg/ml Gentamycin (both Wisent Inc., St-Bruno, QC). Leukemic cells were grown on S17 feeder cells in Iscove's (Invitrogen) with its corresponding cytokines and supplements as for healthy B cells. Clonogenic progenitor assays were performed as has been described by us previously (Fournier et al., 2012; Lebert-Ghali et al., 2010).

Generation of Chimeras

BM chimeras were generated as described before by us (Fournier et al., 2012), with the difference that BM cells from E2A-PBX1/CD3ε−/− transgenic mouse were used in retroviral infections with Hoxa9-GFP or control MSCV-GFP retroviruses. One month posttransplantation mice were killed and GFP positive B cells were sorted from the BM. Transduced B cells were re-transplanted into lethally irradiated C57BL/6 mice together with 2 × 105 total BM cells from congenic B6SJL mice. Peripheral blood analyses were performed bi-weekly following transplantation to monitor leukemia development.

Statistical Analysis

To compare survival curves a logrank Chi-square test was performed using the curve comparison option in GraphPad Prism v.4. All other statistical analysis was done using a Student t-test and F-test.

Clonality Analysis

Genomic DNA was isolated from fresh BM cells using DNAzol (Invitrogen), then digested with NcoI or EcoRI (New England Biolabs, Unit 6 Pickering, ON) and analyzed for retroviral integrations by Southern blotting using a probe against the GFP gene.


We thank in particular Dr. Guy Sauvageau for providing the E2A-PBX1/CD3ε−/− transgenic mice. Martine Dupuis (HMR Research Center) and Danielle Gagné (IRIC) are thanked for sorting of the cells and assistance with flow cytometry analysis. We also acknowledge support from the Quebec ALL (QcALL) biobank and Thomas Sontag for the processing and distribution of leukemia cells. The staff of the HMR animal facility is acknowledged for the great care of the animals. This work was supported by a grant from The Leukemia & Lymphoma Society of Canada. M.F. holds a Scholarship Ph.D. Award of the Cole Foundation. C.E.L.G is a recipient of a Scholarship from the Graduate School Faculty of University of Montréal. The authors declare no competing financial interests.