The TEL-AML1 leukemia fusion gene dysregulates the TGF-β pathway in early B lineage progenitor cells

Induced expression of TEL-AML1 in BaF3 cells.

The GeneSwitch system (20) is based on an autoregulatory feedback loop that involves the binding of a GAL4 regulatory fusion protein, pSwitch, to GAL4 upstream activating sequences in both the promoter controlling expression of the GAL4 regulatory protein and that controlling expression of TEL-AML1. The expression of TEL-AML1 is itself controlled by the presence or absence of the agonist mifepristone, which brings about a conformational change in the pSwitch regulatory protein and its subsequent activation (Supplemental Methods and Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI36428DS1). We established inducible TEL-AML1 in murine BaF3 cells, a putative pro-B cell line (21). Thirty double-positive stable clones expressing the regulatory protein that also showed inducible expression of TEL-AML1 by Western blot and staining via the V5 terminal tag were selected and expanded in liquid culture. Since we observed some degree of cell death in both inducible control and TEL-AML1–expressing clones at the recommended concentrations of mifepristone, we further titrated positive clones against decreasing concentrations of agonist to determine the lowest optimal conditions for protein expression without pleiotropic effects (data not shown). All subsequent experiments were performed with 12.5 pM mifepristone and 1.5 × 10⁴ cells/ml for 3 days (unless otherwise stated). After incubation with mifepristone, predominant nuclear speckled staining was observed by confocal microscopy using an antibody against the V5 tag in all TEL-AML1–expressing cells but not control cells (Figure 1, A and B, and Supplemental Figure 2). The results from one representative clone (i.e., 1/27) are shown in Figure 1B and from another, in Supplemental Figure 3. In a number of different clones, expression of TEL-AML1 was observed by Western blot analysis in as little as 4 hours (Figure 1C and data not shown).

BaF3 cells expressing TEL-AML1 retain IL-3 dependence, grow more slowly than controls, but are less sensitive to the antiproliferative effects of TGF-β.

We next evaluated the growth profile of TEL-AML1–negative and –positive BaF3 cells in standard culture conditions by determining viable cell numbers (Figure 2). For this purpose, 2 TEL-AML1–expressing clones were used (with essentially identical results), and here we show the data for one of them. No difference in growth was detected with inducible control clones grown in the presence of mifepristone (Figure 2A), while expression of the fusion gene caused a slowing of cell growth (Figure 2B). We have also observed, consistently, a slower proliferation rate in several murine B cell progenitor cell lines following transfection with retroviral TEL-AML1 constructs (C. Champseix, S. Tsuzuki, T. Enver, and M. Greaves, unpublished observations). BaF3 cells expressing TEL-AML1 retained dependence on IL-3 for viability and proliferation (Supplemental Figure 4B).

Next, we investigated whether TEL-AML1 expression, despite its negative impact on cell proliferation rate, could, nevertheless, provide some advantage in the context of growth inhibitors or apoptotic stimuli. We first tested the response of the cells to the addition of IFN-γ or TNF, IL-3 deprivation, or serum starvation, but saw no differences due to TEL-AML1 expression (Supplemental Figure 4B and data not shown). There was, however, a consistently different response of TEL-AML1–positive cells to the growth inhibitor TGF-β1 (TGF-β) at 10 ng/ml. As in previous studies (22), parental BaF3 cells were sensitive to the antiproliferative effects of TGF-β over a dose range of 1–100 ng/ml (see Methods and Supplemental Figure 4A), as were inducible control cells (Figure 2A), but TEL-AML1–positive BaF3 cells showed marked resistance to its effects, as evidenced by the finding that TGF-β did not further reduce their slower rate of proliferation (Figure 2B). The net result is that in the presence of TGF-β, TEL-AML1–positive cells proliferate at approximately the same rate as normal, equivalent cells.

In further examination of the inhibitory effects of TGF-β, we pulsed cells with BrdU for 30 minutes and analyzed the cell cycle of this synchronized population of cells after 10 hours of culture in the presence or absence of TGF-β (Figure 2C). TGF-β affected the cell cycle of TEL-AML1–negative cells, reducing the proportion of cells in S plus G₂ relative to M phase (from 49% to 22%). In the absence of the mifepristone inducer, cycling of the TEL-AML1 clone was also reduced by TGF-β (51% to 32%). However, when TEL-AML1 expression was induced, the more slowly cycling cells were again resistant to further proliferative inhibition by TGF-β; i.e., there was no reduction in the proportion of cells in S plus G₂ relative to M phase (Figure 2C; 21% vs. 24%). These data suggested that in the presence of TGF-β, cells expressing TEL-AML1, although proliferating slowly, might acquire at least proliferative parity if not a selective advantage.

We next performed a coculture experiment with a range of percentages of TEL-AML1–negative and –positive cells in the presence or absence of TGF-β and monitored the percentage of TEL-AML1–positive cells. Figure 3 provides data from one of 5 independent experiments that gave consistent results. After 6 days in the absence of TGF-β, there was a large relative increase in the percentage of TEL-AML1–negative cells, in accordance with their faster growth. However, in the same in vitro culture conditions in the presence of TGF-β, 10-fold more TEL-AML1–positive cells were consistently present, maintaining proliferative parity (Figure 3).

The increase in endogenous p27^KIP1 transcription in response to TGF-β is blocked in the presence of TEL-AML1.

In the presence of TEL-AML1 alone, we saw a relative increase in transcript levels of the cyclin-dependent kinase (Cdk) inhibitor proteins p27^KIP1 (CDKN1B) and p21 (CDKN1A) (Figure 4A and Supplemental Figure 5), which control G₁ phase progression and S phase entry (23). This upregulation is concomitant with the change in cell cycle. However, it has also been shown that TGF-β induces G₁ arrest in B cells via upregulation of p27^KIP1 protein (23). We therefore investigated next whether the resistance of TEL-AML1–positive cells to TGF-β–induced growth inhibition was associated with an impact on p27^KIP1 expression. Real-time quantitative PCR (Q-PCR) analysis of p27^KIP1 expression levels was performed in TEL-AML1–positive and –negative cells treated with TGF-β for different periods of time (Figure 4B) and in the absence of TGF-β (Supplemental Figure 5). As anticipated, inducible control cells in the presence of TGF-β showed a progressive increase in p27^KIP1 expression. Conversely, in the presence of TGF-β, TEL-AML1–positive cells exhibited minimal further increase in p27^KIP1 expression. In some experiments (as in Figure 4B), there was a transient increase in p27^KIP1 in the presence of TGF-β and TEL-AML1, but this was not consistently seen.

The expression of TEL-AML1 markedly represses the response of the mouse Igα promoter to TGF-β.

To determine whether TEL-AML1 expression interfered only with the antiproliferative activity of TGF-β or whether its overall signaling was affected, we tested the response to TGF-β of one of the known target gene promoters. Using transient transduction reporter assays with a construct that contained the mouse Igα promoter (positively regulated by TGF-β; ref. 24), we observed that expression of TEL-AML1 markedly repressed the response of this promoter to TGF-β (Figure 4C).

Smad2/3 signaling in the presence of TEL-AML1.

To gain some insight into the possible mechanism underlying the lack of response of TEL-AML1–positive cells to TGF-β, we analyzed the TGF-β pathway of these cells. First we investigated the early steps of the pathway: TGF-β binding to TGF-β II receptor and activation of the receptor complex by Smad2/3 phosphorylation in the presence or absence of a potent and specific inhibitor of TGF-β (SB-431542) (25, 26). By Western blot analysis with anti–p-Smad2/3, we showed that the phosphorylation of endogenous Smad2/3 protein, in response to TGF-β, was unaffected in TEL-AML1–positive cells, inferring that disruption to the TGF-β pathway by TEL-AML1 was subsequent to Smad2/3 activation (Supplemental Figure 6 and data not shown). Furthermore, using Q-PCR array analysis, we also analyzed the expression of a focused panel of genes known to be related to bone morphogenetic protein–mediated TGF-β signal transduction and assessed their status in the presence or absence of TEL-AML1. No changes in the levels of TGF-β II receptor gene expression were observed (data not shown), again consistent with downstream disruption of the pathway.

It has been reported that other leukemic fusion proteins that incorporate AML1, such as AML1/EVI-1 and AML1/ETO, associate with Smad3 (27, 28) and do so via the DNA binding Runt domain (27), or EVI1 in the case of AML1-EVI1 (28). Since this domain is retained in the TEL-AML1 fusion, we investigated whether TEL-AML1 was able to associate with Smad3. Smad3 immunoprecipitation from lysates of TEL-AML1–positive and –negative cells, grown for 48 hours in the presence or absence of TGF-β, confirmed association of TEL-AML1 with Smad3 (Figure 4D). We next used the CAGA₁₂ promoter reporter gene to assess the response to TGF-β in the absence of a binding site for (TEL-)AML1. The CAGA₁₂ promoter contains 12 tandem copies of the DNA binding site for Smad3 alone linked to luciferase (25) but no binding site for AML1. The results indicated reduced activation of the CAGA₁₂ promoter by TGF-β in the presence of TEL-AML1, which cannot bind to this promoter (Supplemental Figure 7A). Furthermore, using EMSA with protein extracts isolated from REH t(12;21) cells and a DNA probe that contains the AML_U and AML_D binding sites as well as SMAD_A and SMAD_B binding sites (29), we show that TEL-AML1 is able to bind to its cognate binding site in the presence of Smad3 (Supplemental Figure 7B). The combined data suggest that the repressive effect seen at this TGF-β–responsive promoter in the presence of TEL-AML1 operates through interaction with Smad3 but may not require full DNA binding of the fusion protein itself. Similarly, the corepressor mSin3A is known to regulate diverse signaling pathways in both normal and malignant cell growth by forming large multiprotein repressor complexes (15, 16). In cells induced to express TEL-AML1, immunoprecipitation with antibody against mSin3a confirmed its interaction with the fusion protein (data not shown).

Functional impact of TEL-AML1 expression in progenitor cells from transgenic mice.

As the lineage affiliation of BaF3 is not unambiguous, we analyzed the effect of TEL-AML1 protein expression in an alternative murine model system: BM-derived B lineage progenitor cells from mice transgenic for TEL-AML1. As expression of the fusion gene was regulated by the IgH enhancer, we anticipated that TEL-AML1 protein would be selectively expressed in the B cell lineage and, possibly, at lower levels in earlier progenitors, in which the enhancer element might be accessible to transcription factors and therefore active (30). Two transgenic lines expressing Eμ TEL-AML1 were generated. In more than 100 mice followed for 18 months, no leukemias developed, commensurate with the view that TEL-AML1 is in itself insufficient to cause overt leukemia (9, 31).

We cloned BM colony-forming cells from mice in the presence of Flt-3 ligand, IL-7, and SCF. Two morphological types of colonies were generated in both wild-type and TEL-AML1 mice (Figure 5A). These distinctive colonies were qualitatively distinguished, and we refer to them as “spread” (or diffuse) and “tight” (or dense). TEL-AML1 mice consistently showed higher numbers of spread colonies and fewer tight colonies compared with wild-type mice (Figure 5B). Replating assays revealed that spread colonies could form both spread and tight secondary colonies, but replated tight colonies only produced tight colonies. This suggests that spread CFCs are developmentally less differentiated than tight colonies, which is compatible with their phenotypes (see below). We next assessed the impact of TGF-β on progenitor CFCs with and without TEL-AML1 expression by selectively picking spread versus tight colonies and replating them in the presence of TGF-β. We observed that both spread and tight TEL-AML1–positive colonies were consistently less sensitive to the inhibitory effects of TGF-β than were wild-type colonies. Reduced sensitivity to TGF-β was reflected in colony numbers (Figure 5C and Supplemental Figure 8; data from 5 experiments). Additionally, residual colonies in TGF-β–treated TEL-AML1–positive cell were usually larger than in wild-type populations (Supplemental Figure 9).

Analysis of cells retrieved from these colonies according to morphology, immunophenotype (Figure 5A and Figure 6), and gene expression (Q-PCR) (Supplemental Figure 10) revealed that the spread colonies were a mixed population of lymphoid and myeloid cells, whereas cells in the tight colonies were all lymphoid. Wild-type and TEL-AML1–positive tight colonies showed a similar pre-B cell phenotype, being B220-, CD19-, and BP1-positive but negative for Flt-3, Sca-1, and CD11b (Figure 6). In contrast, the spread colonies were positive for Flt-3, Sca-1, and c-Kit, indicative of a more immature, progenitor-like phenotype. In line with the morphological variety of cells within the spread colonies, we also observed distinct and separate populations of cells either positive or negative for CD11b (Figure 6). These were flow sorted for future analysis. The CD11b-negative cells derived from TEL-AML1 transgenic mice were exclusively lymphoid in morphology and expressed RAG1, CD79a, and TEL-AML1 (Supplemental Figure 10). The CD11b-positive cells had myeloid morphology, expressed low levels of CD79a but not RAG1, and had low or negligible levels of TEL-AML1 and most closely resemble the pre-pro-B cells of Hardy’s fraction A (32). These data indicate that the spread colonies are derived from a common lympho-myeloid progenitor with (in the assay used) B lineage progeny. We also observed a substantial increase in the number of c-Kit⁺Sca-1⁺ cells in spread colonies in the presence of TEL-AML1 expression, which is in keeping with previous findings in BM transplantation models with TEL-AML1 (6, 8) and indicative of a selective impact on both multipotent (lympho-myeloid) and very early B cell progenitor cells.

TEL-AML1 expression provides selective advantage to candidate “preleukemic” stem cells in the presence of TGF-β.

Human cord blood CD34⁺ cells transduced with TEL-AML1 via a lentiviral vector produce a preleukemic-like phenotype in NOD/SCID mice (13). This is dependent upon the expansion of a putative pre-LSC stem cell population with the immunophenotype CD34⁺CD38^–CD19⁺ (13). We have assessed whether abrogation of TGF-β sensitivity by TEL-AML1 protein in murine models can be replicated in a system that more approximates clinical preleukemia in patients and, in particular, whether the candidate stem cell can be selected. Candidate pre-LSCs were generated in vitro by culture of TEL-AML1–transduced CD34⁺CD38^–CD19^– cord blood cells on MS-5 stroma (13). From an initial average inoculum of 93.3 TEL-AML1–expressing (GFP-positive) cells (SD 15.3, SEM 8.82, range 80–110), 5,146.7 CD34⁺CD38^–CD19⁺ (candidate pre-LSCs) were generated (SD 147, SEM 85.1, range 4,980–5,260). These cells were then replated on MS-5 stroma in the presence of TGF-β. After an additional 3 weeks, they had proliferated to give rise to 19,667 pre-LSCs (SD 4,400.4, SEM 2,540, range 15,300–24,100). Thus, in the presence of TGF-β, there is an absolute increase in the number of TEL-AML1–expressing pre-LSCs of approximately 4-fold (average 3.82, range 3.1- to 4.6-fold). Interestingly, this net increase in pre-LSCs that occurs in the presence of TGF-β was accompanied by an absolute decrease in cell number of more mature TEL-AML1–expressing and wild-type cells, and this was not seen in the absence of TGF-β. The combination of increased numbers of pre-LSCs and decreased numbers of mature cells resulted in the 12-fold proportionate increase in pre-LSCs presented in Figure 7.

Cell culture and TEL-AML1 expression.

All BaF3 clones were cultured in RPMI medium supplemented with 10% FCS, 2% MoIL-3–conditioned medium containing IL-3 (56), 10 μM 2-mercaptoethanol, and 0.2 mg/ml hygromycin B (Invitrogen). In addition inducible clones were cultured in the presence of 0.05 mg/ml Zeocin (Invitrogen) (see Supplemental Methods).

The expression of TEL-AML1 was induced by adding to the culture medium 0.0125 nM mifepristone (Invitrogen)/3.8 × 10⁵ cells for 3 days and then in 0.01 nM mifepristone to maintain long-term expression of TEL-AML1. Efficiency of TEL-AML1 induction was verified by immunocytochemistry, flow cytometry, and Western blot analysis using an anti-V5 antibody (Invitrogen). Hereafter, control clones are referred to as “inducible control” and test clones as “inducible TEL-AML1.”

Analysis of TGF-β effects on cell growth profiles.

The BaF3 clones (2.5 × 10⁴) were grown in their standard culture conditions and in the presence or absence of 10 ng/ml rhTGF-β (R&D Systems). In pilot experiments, proliferation of parental BaF3 cells was inhibited by about 90% over the dose range 1–100 ng/ml (Supplemental Figure 4A). We elected to use a concentration of 10 ng/ml in all subsequent experiments. Some experiments were also carried out in the presence of the TGF-β inhibitor SB-43154 (26), a gift of Caroline Hill (Cancer Research UK, London Research Institute, London, United Kingdom). Cell growth was measured at the indicated incubation times by determining the number of viable cells, using trypan blue dye exclusion.

Cell cycle analysis.

BaF3 clones with and without 3-day induction of TEL-AML1 expression were incubated with 20 μM BrdU (Sigma-Aldrich) for 30 minutes in standard culture conditions. Cells were washed and cultured in fresh medium, without BrdU, in the presence or absence of 10 ng/ml rhTGF-β and 0.01 nM mifepristone. After 10 hours, cells were harvested, and 1 × 10⁶ cells were fixed with ice-cold 70% ethanol for 30 minutes at 4°C, washed with PBS, and treated with 0.1% pepsin (Sigma-Aldrich) in 2 M HCl for 20 minutes at room temperature. The acid was neutralized with 0.1 M Na₂B₄O₇×10H₂O, pH 8.5. Cells were stained with FITC-conjugated anti-BrdU antibody (BD) for 30 minutes at room temperature. Samples were resuspended in 10 μg/ml propidium iodide (Sigma-Aldrich) and 100 μg/ml RNase (Sigma-Aldrich) in PBS for 20 minutes at 37°C and then analyzed by flow cytometry.

Coculture experiment.

After induction of TEL-AML1 for 3 days, clones were cocultured with negative control clones (at a ratio of about 85% to 15%) in RPMI medium with 10% FCS, 2% MoIL-3 conditioned medium, 10 μM 2-mercaptoethanol, 0.2 mg/ml hygromycin B, and 0.01 nM mifepristone and in the presence or absence of 10 ng/ml rhTGF-β. The percentage of TEL-AML1–positive cells was measured by flow cytometry after staining for intracellular V5 as described in Supplemental Methods.

Transgenic mice.

Lines of TEL-AML1 transgenic mice were generated by insertion of a full-length TEL-AML1 cDNA into the TOK4 expression vector (a gift of Michael Elverstein, Cambridge Laboratory of Molecular Biology, Medical Research Council, Cambridge, United Kingdom) and injection into blastocysts of B6CBAF-1 mice. The construct was driven by the human β-globin promoter and lymphoid lineage specificity achieved via the immunoglobulin (IGH) heavy chain enhancer. All experiments were approved by and conform to the standards of the animal use ethics committee, the Institute of Cancer Research, London, United Kingdom, and Home Office license requirements (PPL 70/5949). Mice were genotyped by PCR using previously published primers (46). Nontransgenic littermates were used as age-matched wild-type controls.

Isolation of human hematopoietic progenitor cells.

Cord blood from anonymized healthy donors was purchased from either the UK National Blood Service or from StemCell Technologies (CB003F) in accordance with local ethical procedures (NHS Oxfordshire Local Research Ethics Committee approval 04/Q1605/22). Anonymized peripheral blood and BM samples were from routine clinical specimens taken from patients at Great Ormond Street Hospital, London. Fully informed written parental consent was obtained in accordance with national guidelines. Total mononuclear cells (MNCs) were isolated by Ficoll gradient centrifugation. CD34⁺ cell fractionation from cord blood was achieved initially by magnetic bead separation (StemCell Technologies or Miltenyi Biotec). CD34-enriched cells from cord blood or MNCs from peripheral blood and BM were stained with anti-CD19–APC (Dako), -CD34-PE (BD Biosciences — Pharmingen), and -CD38-FITC (Dako).

Preparation of lentivirus, hematopoietic progenitor cell transduction, and identification of pre-LSCs.

The previously described TEL-AML1 myc-tag fusion construct (6) was cloned into the pHR-SINCSGW lentivirus (57), which carries an emerald-GFP reporter. Lentiviruses were pseudotyped with the vesicular stomatitis virus G (VSVG) protein by transient transfection of 293T cells. High-titer viral stocks were prepared by ultracentrifugation (58). Cord blood hematopoietic progenitor cells were infected for 24 hours by lentivirus (either TEL-AML1 or control virus) with an MOI of 100 (or 5 × 10⁷ IU/ml for small numbers of cells) in serum-free medium (StemSpan; StemCell Technologies) supplemented with SCF (100 ng/ml), Flt-3 ligand (100 ng/ml), TPO (20 ng/ml), and IL-6 (20 ng/ml) (all growth factors were from Peprotech) (59, 60). Transduced cells were inoculated onto MS-5 cells. Candidate pre-LSCs were defined as cells expressing TEL-AML1 (i.e., GFP-positive) with the immunophenotype CD34⁺CD38^–CD19^– (13).

MS-5–supported culture.

Lentivirus-transduced CD34⁺ cord blood cell populations were maintained on MS-5 stromal cells. The latter cells were maintained in α-MEM medium complemented with 2 mM l-glutamine, 2 mM sodium pyruvate, 10% FBS (StemCell Technologies). After addition of transduced CD34⁺ cells, SCF and G-CSF (Peprotech) were added, and weekly half-medium changes were performed for the duration of the culture (61, 62) (further details are given in ref. 13).