BCL6-mediated repression of p53 is critical for leukemia stem cell survival in chronic myeloid leukemia

To study genes potentially contributing to the maintenance of CML cells exposed to TKI treatment, CML cells were incubated in the presence or absence of the TKI Imatinib and subjected to gene expression analysis. Because Stat5 represents a central mediator of BCR-ABL1 signaling, we studied a BCR-ABL1 leukemia mouse model in the context of inducible deletion of Stat5 (Fig. S2). This analysis showed that many TKI-induced gene expression changes, including BCL6, are in fact Stat5-dependent (Fig. 1 A). TKI-induced gene expression changes that occurred in a Stat5-independent manner involved multiple erythroid lineage transcripts, including hemoglobins (HBA1, HBB, HBD, and HBE1), erythroid surface antigens (CD36 and GYPA), Beatty’s protein (elliptocytosis; EPB41), and δ-aminolevulinate synthase (ALAS2; Fig. 1 A). TKI-induced up-regulation of BCL6 in human CML cells was confirmed in vitro at the mRNA and protein levels (Fig. 1 B and D), and other gene expression changes were confirmed by quantitative RT-PCR (Fig. S3). In agreement with in vitro observations, BCL6 is strongly up-regulated in CML cells from patients who were treated with TKI.

CML cells were isolated via leukapheresis from two patients in CP, and then sorted for CD34⁺ CD38⁻ and CD34⁺ CD38⁺ multilineage progenitors and more mature CD34⁻ CD38⁻ and CD34⁻ CD38⁺ transient amplifying cells. Treatment with Imatinib for 12 h resulted in strong up-regulation of BCL6 mRNA levels in the CD34⁺ CML cells, but not in CD34⁻ CML cells, regardless of CD38 expression (Fig. 1, E and F). These findings indicate that the ability to up-regulate BCL6 in response to TKI treatment is restricted to cells within the pool of CD34⁺ cells with multilineage potential.

Previous work implicated FoxO factors as positive regulator of BCL6 (Fernández de Mattos et al., 2004). In agreement with this study, we found that FoxO activity is required for TKI-induced BCL6 expression in CML cells (Fig. 1 G). Although AKT-mediated phosphorylation downstream of the BCR-ABL1 kinase results in global inactivation of FoxO factors (Tran et al., 2002; Fig. S1), the Pten phosphatase is required for FoxO activation. Here, we demonstrate that conditional deletion of Pten abrogates the ability of CML-like cells to up-regulate BCL6 in response to TKI treatment (Fig. 1 G). In fact, overexpression of a constitutively active FoxO3A mutant was sufficient to induce an ∼10-fold increase of BCL6 mRNA levels in human CML cells (Fig. 1 H). The finding of FoxO3A as upstream regulator of BCL6 is of particular importance, given that FoxO3A was recently identified as a requirement for the maintenance of CML-initiating cells (Naka et al., 2010).

We then tested the significance of TKI-induced BCL6 in a genetic loss-of-function experiment. To this end, BCL6^+/+ and BCL6^−/− bone marrow hematopoietic progenitor cells were transformed with p210 BCR-ABL1 according to a classical model for CML in mice (Pear et al., 1998; Li et al., 1999). The transformation efficiency of BCL6^+/+ and BCL6^−/− hematopoietic progenitor cells was undistinguishable and, in both cases, after ∼1 wk a growth factor–independent CML-like leukemia developed (Fig. S4; n = 3). Because BCR-ABL1 can give rise to both B cell lineage and myeloid lineage leukemia (i.e., CML), a myeloid-specific transformation protocol was used in all experiments (Li et al., 1999). Myeloid lineage identity of the BCR-ABL1–transformed cells was routinely verified by flow cytometry. The phenotype of CML-like cells was characterized in depth when the CML-like leukemia model for BCL6^+/+ and BCL6^−/− hematopoietic progenitor cells was established (Fig. S5 and Fig. S6).

Because the pool of CD34⁺ cells with multilineage potential represents the main source of BCL6 expression in human CML (Fig. 1, E and F), we studied the effect of BCL6 deficiency in the equivalent population (Lin⁻ Sca-1⁺ c-Kit⁺ [LSK]) in our CML-like mouse model. Given that BCL6 was strongly up-regulated in response to Imatinib-treatment of CML cells (Fig. 1), we tested whether BCL6 regulates sensitivity of CML cells to Imatinib. Mouse LSK cells in CML-like leukemia are highly refractory to Imatinib, and Imatinib concentrations of >3 μmol/liter are required to induce cell death in LSK⁺ CML cells. BCL6-deficient CML LSK cells, however, were sensitive to Imatinib even at concentrations <0.5 μmol/liter (Fig. 2 A). Comparing the response of CML-like cells to Imatinib treatment (10 μmol/liter), BCL6^+/+ CML-like cells were significantly less sensitive compared with BCL6^−/− CML-like cells (mean viability 71.8 ± 2.7% compared with 31.4 ± 1.6% viable cells; P < 0.0003; Fig. 2 B). We next studied whether inducible reconstitution of BCL6 rescues Imatinib-resistance in BCL6^−/− CML-like cells. To this end, BCL6^−/− CML-like cells were transduced with a 4-hydroxy-tamoxifen (4-OHT)–inducible BCL6-ER fusion molecule, which is activated within minutes after 4-OHT addition (Shaffer et al., 2000). Although an ER empty vector control had no effect on the survival of leukemia cells, inducible BCL6 activation conferred a strong survival advantage as reflected by a 30-fold increase of BCL6-ER–transduced cells (Fig. 2, C and D).

Because the aforementioned genetic experiments indicate a critical role for BCL6 in Imatinib resistance of CML cells, we next tested if the implicated synthetic lethality could be targeted pharmacologically using a combination of Imatinib and a BCL6-peptide inhibitor in human CML cells. To this end, human CML cells were incubated in the presence or absence of Imatinib, a novel retro-inverso BCL6 peptide inhibitor (RI-BPI; Cerchietti et al., 2009), or a combination of both (Fig. 2 E). Although RI-BPI alone did not significantly affect CML cell viability, it strongly enhanced the effect of Imatinib, except for one case, in which nearly all cells underwent apoptosis upon Imatinib treatment alone (Fig. 2 E).

When mouse LSK cells were transformed with BCR-ABL1 in the presence of IL-3, IL-6 and SCF, the vast majority of BCL6^+/+ CML-like cells retain an LSK-phenotype. In contrast, the LSK population of the BCL6^−/− CML-like cells rapidly undergoes apoptosis. Within 2 wk, the LSK population, which is thought to comprise the pool of LICs was reduced from ∼15 to <1% in BCL6^−/− CML (Fig. 3, A and B). As opposed to LSK cells, the non-LSK population consistently lacks the ability to initiate leukemia in serial transplantation experiments (Hu et al., 2006; Neering et al., 2007; Ito et al., 2008; Zhao et al., 2009). Besides loss of the LSK phenotype, an Affymetrix GeneChip expression analysis revealed loss of expression of multiple other stem cell–related molecules in BCL6^−/− CML-like cells (e.g., Egr1, Ptgs1, Slamf1/CD150, Gfi1, Rora, Abcg2; Fig. S7).

These changes may occur for various reasons, such as enhanced differentiation, reduced self-renewal, or selective apoptosis/depletion of BCL6^−/− LICs. To test these possibilities, we sorted viable LICs (LSK⁺ phenotype) and transient amplifying cells (LSK⁻ phenotype) from freshly generated BCL6^+/+ and BCL6^−/− CML-like leukemia. At this time, BCL6^−/− CML-like leukemia still had a high frequency of LSK⁺ cells. For each population, between 100,000 and 1 million cells were sorted with >98% viable cells. After 20 h of incubation, cell counts moderately increased for transient amplifying cells from both BCL6^+/+ and BCL6^−/− CML-like leukemia. Also counts for LSK⁺ LICs from BCL6^+/+ CML-like leukemia increased, whereas counts for LSK⁺ LICs from BCL6^−/− CML-like leukemia significantly dropped (Fig. S8). Selective reduction of LSK⁺ LIC counts from BCL6^−/− CML-like leukemia is consistent with LIC depletion observed in Fig. 3, A and B. Strikingly, flow cytometry revealed that the majority of LSK⁺ LICs from BCL6^−/− CML-like leukemia was preapoptotic after 20 h. These findings show that BCL6 represents a critical factor for LIC survival in CML. In the absence of BCL6 function, LICs are poised to undergo apoptosis.

To identify transcriptional targets of BCL6 in human CML cells, we performed a genome-wide mapping analysis of BCL6 recruitment using ChIP-seq. Because BCL6 is known to function as its own repressor (Mendez et al., 2008), we studied recruitment of BCL6 to the BCL6 and HPRT promoters as positive and negative controls, respectively (Fig. 3 C). Several molecules in the DNA damage/checkpoint signaling pathway, including CHEK1, CARF (CDKN2AIP), p53, GADD45A, and p21 (CDKN1A) exhibit robust recruitment of BCL6 (Fig. 3 C). ChIP-seq was not informative for ARF, because the JURL CML cells studied carry biallelic deletions at 9p21, including the CDKN2A locus (Fig. 3 C). CARF (CDKN2AIP), which is required for protein stability of ARF (CDKN2A; Hasan et al., 2002), and p53 are both direct transcriptional targets of BCL6 (Fig. 3 C). Therefore, we tested whether the ARF/p53 pathway is deregulated in BCL6^−/− CML-like cells. Indeed, protein levels of both ARF and p53 were increased in the absence of BCL6 (Fig. 3 D). Hence, BCL6 may curb excessive expression of ARF/p53 in CML-like cells.

Because components of the Arf/p53 pathway are mediators of cellular senescence and negatively regulate self-renewal in normal hematopoiesis and leukemia (Oguro et al., 2006), we next tested whether the unrestrained expression of Arf/p53 in the absence of BCL6 interferes with CML cell self-renewal. To this end, BCL6^+/+ and BCL6^−/− LSK cells from CML-like leukemia were plated in semisolid agar and colonies were counted 22 d later. Immediately before plating, viability of cells was verified by flow cytometry (>90%). Strikingly, in the absence of BCL6, LSK cells from CML-like leukemia nearly entirely lack the ability to form colonies, and on most plates not a single colony was detected (Fig. 3 E). In the absence of BCL6, colony formation was reduced by >300-fold, which demonstrates a critical role of BCL6 in self-renewal of CML-initiating cells. Unlike for LICs in CML, BCL6 was dispensable for colony formation in normal LSK. Although normal BCL6^−/− LSK cells had a slightly lower colony count compared with BCL6^+/+ LSK cells, colony formation was not compromised in the absence of BCL6 (Fig. S9). Likewise, a recent study demonstrated that normal LSK cells from BCL6-deficient mice have multilineage potential similar to their BCL6^+/+ counterparts (Broxmeyer et al., 2007). More detailed experiments to address a potentially unrecognized function of BCL6 in normal LSK cells are currently under way.

Because Arf/p53 signaling also affects cell cycle check points, we studied the consequences of BCL6-deficiency on cell cycle regulation in CML-like cells (Fig. 3 F). This analysis revealed a striking anomaly in BCL6^−/− CML-like cells, which exhibit a large subpopulation with “mitotic crisis” phenotype (Fig. 3 F; Li and Dang, 1999). BrdU incorporation showed arrest in G1/S-phase with incomplete DNA replication. Consistent with the high level of DNA damage stress and Arf/p53 activation, LSK⁺ BCL6^−/− CML cells do not survive over longer periods of time. As shown in Fig. 3 (A and B), >95% of LSK⁺ BCL6^−/− CML cells die within 3 wk of leukemic transformation. In cell culture experiments, we observed that BCL6^−/− CML-like cells occasionally ceased to propagate and the entire population underwent cell cycle arrest. In the context of the “mitotic crisis” phenotype (Fig. 3 F), we hypothesized that BCL6^−/− CML-like cells may undergo replicative senescence owing to telomere shortening after multiple cell divisions. Measurement of telomere lengths in BCL6^+/+ and BCL6^−/− CML-like cells by flow FISH (Fig. S10), however, showed that telomere lengths were similar in BCL6^+/+ (31.4 ± 1.1 kb) and BCL6^−/− CML-like cells (35.5 ± 0.9 kb).

Because BCL6 functions as a transcriptional repressor of p53 in human and mouse CML cells (Fig. 3, C and D), we hypothesized that BCL6-mediated repression of p53 represents a key element in BCL6-dependent self-renewal signaling. To test this hypothesis, we transduced p53^+/+ and p53^−/− CML-like cells with a 4-OHT–inducible dominant-negative mutant of BCL6 (DN-BCL6-ER^T2; Shaffer et al., 2000). This form competes with wild-type BCL6 for DNA binding, but lacks the BCL6-BTB (Bric-à-Brac, Tramtrack, and Broad complex) domain and thus lacks the ability to function as transcriptional repressor. 4-OHT–mediated induction of DN-BCL6-ER^T2 nearly completely suppressed colony formation in p53^+/+ CML-like cells, whereas colony formation in p53^−/− CML-like cells was only slightly reduced (Fig. 4 A). Likewise, induction of DN-BCL6-ER^T2 induced G1/0 cell cycle arrest in p53^+/+, but not p53^−/− CML-like cells (Fig. 4 B). In contrast to p53^−/− CML-like cells, 4-OHT–mediated induction of DN-BCL6-ER^T2 resulted in a drastic growth disadvantage in p53^+/+ CML-like cells (reduction of transduced cells to <20% within 5 d; Fig. 4, C and D). These findings suggest that BCL6-mediated repression of p53 is not only required for the initiation of CML colonies (Fig. 4 A), but also for the proliferation and survival of CML LICs (Fig. 4, B and D).

To formally test the requirement of BCL6 for leukemia initiation in vivo, we performed a classical SCID leukemia-initiating cell (SL-IC) assay (Bonnet and Dick, 1997). 500,000 luciferase-labeled BCL6^+/+ and BCL6^−/− CML-like leukemia cells (Lin⁻ Sca-1⁺ c-Kit⁺ CD13⁺; Fig. S5) were injected intrafemorally into NOD/SCID mice, and leukemia initiation and expansion was monitored by bioimaging (Fig. 5 A). Intrafemoral injection was chosen to focus on leukemia initiation and to exclude potential defects in bone marrow homing and engraftment of BCL6^−/− CML-like cells as confounding variables (Krause et al., 2006). Although BCL6^+/+ LSK cells from CML-like leukemia rapidly initiated leukemia, bioimaging revealed persistence of BCL6^−/− LSK cells at the site of injection, but no overt leukemia initiation (Fig. 5 A). Our intention was to perform serial transplantation experiments to assess leukemia initiation potential of BCL6^−/− CML-like leukemia in secondary or tertiary transplants. Because BCL6^−/− LSK cells failed to initiate leukemia in primary transplant recipients, serial transplantation was not possible. Consistent with our colony formation assays, the defect of leukemia-development from BCL6^−/− CML-like cells reflects failure to maintain leukemia-initiating cells, as exemplified by the progressive loss of the LSK⁺ population in BCL6^−/− CML-like leukemia (Fig. 3, A, B, and E).

We next tested whether loss-of-function of BCL6 also affects human CML cells. To this end, human CML cells (KCL22 cell line) were transduced with DN-BCL6-ER^T2 or the ER^T2 empty vector control, labeled with lentiviral firefly luciferase, and injected into sublethally irradiated NOD/SCID mice (Fig. 5 B). After visible engraftment of CML, NOD/SCID mice were treated with intraperitoneal injections of 4-OHT over 3 intervals of 10 d (20 mg/kg 4-OHT/per day). Consistent with our in vitro observation, inducible inhibition of BCL6 compromised leukemia initiation in vivo (Fig. 5 B) and significantly prolonged overall survival of xenografted NOD/SCID mice (Fig. 5 C). We conclude that BCL6 function is critical for initiation of human CML in vivo.

We next tested whether pharmacological inhibition of BCL6 (RI-BPI) can interfere with leukemia initiation in human CML. We performed a colony formation assay with human CML cells that were plated on semisolid agar plates with vehicle or 5 μmol/liter RI-BPI. Consistent with our observation that BCL6^−/− CML-like cells nearly entirely lacked the ability to form colonies (Fig. 3 E), RI-BPI–mediated inhibition of BCL6 reduced colony numbers by three- to eightfold compared with untreated cells in a BCL6^+/+ context (Fig. 5 D). More importantly, RI-BPI also interfered with leukemia initiation of human CML cells in vivo. Upon treatment with RI-BPI, xenografted human CML cells (KCL22 cell line) failed to initiate leukemia in transplant recipients. Treatment of human CML cells with RI-BPI increased overall survival of recipient mice and latency of leukemia. 3 of 10 mice that were xenografted with human CML cells (KCL22 cell line) in the RI-BPI group were sacrificed after 180 d (Fig. 5, E and F) and had no indication of leukemia (no CML cells detectable in bone marrow and spleen), compared with no such cases in the untreated controls.

To study the effect of acute BCL6 inactivation in patient-derived CML cells, we incubated CML cells from 5 patients in CP (CP22-CP26; Table S1) and 1 patient in blast crisis (BC12; Table S1) in 5 μmol/liter RI-BPI or Vehicle for 2 h. After this incubation period, RI-BPI was washed out and cells were cultured in the presence of 100 ng/ml SCF, 100 ng/ml G-CSF, 20 ng/ml FLT3, 20 ng/ml IL-3, and 20 ng/ml IL-6. In one set of experiments, cells were stained with CFSE to track cell divisions over time. Flow cytometry revealed that RI-BPI–treated CML-CP samples selectively lost CD34⁺ CD38⁻ LICs (Fig. 6, A and B), whereas CD34⁻ subpopulations remained intact (Fig. 6 C). Likewise, RI-BPI caused cell cycle arrest in CD34⁺ CD38⁻ LICs (Fig. 6 B), whereas other subpopulations continued to divide as measured by CFSE dye dilution. In one case of blast crisis CML, RI-BPI neither affected the CD34⁺ CD38⁻ LIC population nor induced cell cycle arrest in CD34⁺ CD38⁻ LICs. These findings are in agreement with cell cycle deregulation of BCL6^−/− CML-like cells and suggest that acute inhibition of BCL6 function can commit CD34⁺ CD38⁻ LICs to eradication in CML-CP, but not in CML-BC (Fig. 6 C).

A recent work identified a critical role for FoxO factors in the maintenance of leukemia-initiating cells in CML (Naka et al., 2010). In this study, we demonstrate that BCL6 functions as key effector molecule downstream of FoxO and prevents CML LIC depletion by transcriptional repression of Arf/p53. Although Imatinib successfully induces cell death in transient amplifying CML cells, it fails to eradicate leukemia-initiating cells in CML. In fact, both FoxO activity (Naka et al., 2010) and BCL6 expression levels (Fig. 1) are increased by Imatinib, which provides a rationale for selective survival of leukemia-initiating cells in CML during long-term Imatinib therapy. FoxO factors are critical for LIC maintenance in CML-CP, but not CML-BC (Naka et al., 2010), and BCL6 inhibition leads to the eradication of LICs in CML-CP, but not CML-BC (Fig. 6, A–C). Consistent with these findings, both FoxO3A and BCL6 expression levels are tightly correlated during progression of CML-CP toward CML-BC (Fig. S11). We propose that pharmacological inhibition of BCL6 breaks the quiescence program of LICs in CML and renders them vulnerable to drug treatment, e.g., through derepression of MYC (Fig. S12) and reactivation of the Arf/p53 pathway (Fig. 3 D and Fig. S1). Targeting cellular quiescence for selective leukemia stem cell eradication was recently proposed based on observations on IFN-α–, G-CSF–, or As₂O₃-mediated activation (“awakening”) of dormant LICs (Essers and Trumpp, 2010; Trumpp et al., 2010).

Here, we show that pharmacological inhibition of BCL6 leads to LIC apoptosis (Fig. S8) and effectively prevents leukemia initiation from xenografted human CML cells in vivo. Based on these findings, we propose a dual targeting strategy, in which tyrosine kinase inhibitors (e.g., Imatinib) to target the transient amplifying pool of CML cells are coupled with BCL6 inhibition that will target quiescent LICs. Pharmacological options include the BCL6 peptide inhibitors used here (Cerchietti et al., 2009) or newly developed small molecule inhibitors against BCL6 (Cerchietti et al., 2010). Pharmacological inhibition of BCL6, thus, represents a fundamentally novel strategy to eradicate LICs in CML. Clinical validation of this concept could limit the duration of TKI treatment in CML patients, which is currently life-long, and potentially decrease the risk of blast crisis transformation.

Patient samples (Table S1) were provided from the German CML Study Group (in compliance with the internal review board of the University of Southern California Health Sciences Campus, the University of California San Francisco, the Oregon Health and Science University, and the Universität Heidelberg Klinikum Mannheim). In total, 26 cases of CP CML and 12 cases of CML blast crisis were studied, and the characteristics of these patients are summarized in Table S1 (CP1-CP26 and BC1-BC12). The human CML cell lines EM2, JURL-MK1, KCL22, KU812, KYO, LAMA84, MEG1, and MOLM6 were obtained from the German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany. Human leukemia cells were maintained in RPMI-1640 (Invitrogen) with GlutaMAX containing 20% fetal bovine serum, 100 IU/ml penicillin, and 100 µg/ml streptomycin at 37°C in a humidified incubator with 5% CO₂.

For all retroviral transductions with BCR-ABL1, the myeloid-restricted protocol described in Li et al. (1999) was used, which results in CML-like disease. In brief, bone marrow cells were cultured either in Iscove’s modified Dulbecco’s medium (IMDM; Invitrogen) with GlutaMAX containing 20% fetal bovine serum, 100 IU/ml penicillin, 100 µg/ml streptomycin, 50 µM 2-mercaptoethanol, 10 ng/ml recombinant mouse IL-3, 25 ng/ml recombinant mouse IL-6, and 50 ng/ml recombinant mouse SCF (PeproTech) or retrovirally transformed by BCR-ABL1. Lineage identity of CML-like leukemia cells was authenticated by flow cytometry and quantitative RT-PCR (Fig. S5) and microarray analysis (Fig. S6). Mouse CML-like cells and human CML cells were then labeled with lentiviral firefly luciferase (D.B. Kohn, University of California Los Angeles, Los Angeles, CA; Table S2), selected based on antibiotic resistance (puromycin), and injected either via intrafemoral or intravenous tail vein injection into sublethally irradiated (300 cGy) NOD/SCID recipient mice. Engraftment was monitored using luciferase bioimaging (VIS 100 bioluminescence/optical imaging system; Xenogen). D-Luciferin (Xenogen) dissolved in PBS was injected intraperitoneally at a dose of 2.5 mg per mouse 15 min before measuring the light emission.

Retroviral constructs encoding BCR-ABL1 (R. Van Etten, Tufts University, Boston, MA), FoxO3A (D.A. Fruman, University of California Irvine, Irvine, CA), dominant-negative BCL6 (DN-BCL6-ER^T2), inducible activation of BCL6 (BCL6-ER^T2; A.L. Shaffer and L.M. Staudt, National Cancer Institute, Bethesda, MD), Cre-GFP, and empty controls, as well as lentiviral vectors encoding firefly luciferase, were used in transduction experiments as described previously (Duy et al., 2010). A detailed list of vectors and transduction conditions is given in Table S2.

Total RNA from cells was extracted using RNeasy isolation kit from QIAGEN. cDNA was generated using a poly(dT) oligonucleotide and the SuperScript III Reverse transcription (Invitrogen). Quantitative real-time PCR was performed with the SYBRGreenER mix (Invitrogen) and the ABI7900HT real-time PCR system (Applied Biosystems) according to standard PCR conditions. Primers for quantitative RT-PCR are listed in Table S3.

Bone marrow from BCL6^−/− (R. Dalla-Favera, Columbia University, New York, NY; Ye et al., 1997; Table S5) and Stat5ab^fl/fl mice (L. Hennighausen, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD; Liu et al., 1997; Table S5) was harvested and hematopoietic progenitor cells were propagated as described above. Deletion of Stat5a/b^fl/fl, P53^fl/fl, and Pten^fl/fl was induced by retroviral transduction with Cre-GFP (using a GFP empty vector control). Cells were cultured at 37°C in a humidified incubator with 5% CO₂. All mouse experiments were subject to institutional approval by Children’s Hospital Los Angeles IACUC.

Cells were lysed in CelLytic buffer (Sigma-Aldrich) supplemented with 1% protease inhibitor cocktail (Thermo Fisher Scientific). 25 µg of protein mixture per sample were separated on NuPAGE (Invitrogen) 4–12% Bis-Tris gradient gels and transferred on PVDF membranes (Immobilion; Millipore). For the detection of mouse and human proteins by Western blot, primary antibodies were used together with the WesternBreeze immunodetection system (Invitrogen). The following antibodies were used: human BCL6 (clones D8 and N3; Santa Cruz Biotechnology, Inc.), mouse BCL6 (rabbit polyclonal; Cell Signaling Technology), Arf (4C6/4; Cell Signaling Technology), p53 (1C12; Cell Signaling Technology), and pan-specific anti-phosphotyrosine (4G10; Millipore). Antibodies against β-actin were used as a loading control (H4; Santa Cruz Biotechnology, Inc.).

Antibodies against mouse CD44 (IM7), c-kit (2B8), and respective isotype controls were purchased from BD. Anti–mouse Sca-1 antibody (clone 177228) was obtained from R&D Systems. For apoptosis analyses, annexin V, propidium iodide, and 7-aminoactinomycin D (7AAD) were used (BD). Antibodies against human CD34 (563), CD38 (HITZ), and CD90 (OX-7), as well as respective isotype controls, were purchased from BD.

The methylcellulose colony-forming assays were performed with 100,000 normal LSK cells, and BCR-ABL1–transformed mouse CML-like cells or 10,000 human CML cells. Cells were resuspended in mouse MethoCult medium (StemCell Technologies) and cultured on 3-cm diam dishes, with an extra water supply dish to prevent evaporation. After 7–22 d, colony numbers were counted.

Human CML cells were labeled with Hoechst 33342 according to the protocol from Goodell et al. (1996). In brief, cells were suspended in DME (2% FBS and 10 mmol/liter Hepes buffer) and Hoechst 33342 dye was added (5 µg/ml, 90 min at 37°C). After the incubation with Hoechst 33342, cells were centrifuged and resuspended in cold Hanks’ balanced saline solution (2% FBS, 10 mM Hepes buffer, room temperature). Propidium iodide was added to the cells before analysis.

Primary human CML cells were labeled with CFSE (Invitrogen) according to the manufacturer’s protocol. Cells were resuspended in prewarmed PBS (0.1% BSA) and incubated with 0.5 µM CFSE for 10 min at 37°C. Subsequently, 5 volumes of ice-cold PBS were added and cells were incubated for 5 min on ice. After the incubation, cells were washed twice with cold PBS and then cultured in IMDM, 20% BIT, 100 IU/ml penicillin, 100 µg/ml streptomycin, 25 µmol/liter β-mercaptoethanol, 100 ng/ml SCF, 100 ng/ml G-CSF, 20 ng/ml FLT3, 20 ng/ml IL-3, and 20 ng/ml IL-6) for a maximum of 5 d. CFSE levels were measured by flow cytometry together with staining for CD34 and CD38 surface marker expression.

Biotinylated cRNA was generated and fragmented according to the Affymetrix protocol and hybridized to either a U133A 2.0 human, 430 mouse, or Mouse Gene 1.0 ST microarrays (Affymetrix). After scanning (GeneChip Scanner 3000 7G; Affymetrix), the generated data files were imported to BRB Array Tool and processed using the robust multi-array average algorithm. To determine relative signal intensities, the ratio of intensity for each sample in a probe set was calculated by normalizing to the mean value of grouped samples. Microarray data are available from the Gene Expression Omnibus (GEO) under accession nos. GSE24814 (STAT5-deletion; BCR-ABL1–transformed prep cells), GSE24813 (BCL6^+/+ and BCL6^−/− BCR-ABL1–transformed CML-like cells), GSE20987 (BCL6^+/+ and BCL6^−/− BCR-ABL1–transformed pre–B ALL cells), and GSE24493 (Imatinib treated and nontreated human CML cell lines).

Supplementary information for this study includes information on CML patient samples (Table S1), retroviral and lentiviral vectors (Table S2), oligonucleotides used (Table S3), telomere length measurement protocol (Table S4), and an overview of genetic mouse mutants studied (Table S5). Fig. S1 presents a schematic of BCL6-dependent LIC-survival signaling in CML. Validation of gene expression changes were performed in Figs. S2 and S3. Fig. S4 shows transformation efficiency of BCL6^+/+ and BCL6^−/−hematopoietic progenitor cells. Phenotypic authentication of CML-like leukemia by flow cytometry and RT-PCR and microarray analysis is demonstrated in Figs. S5 and S6, respectively. Fig. S7 describes differential gene expression and side population phenotypes in BCL6^+/+ and BCL6^−/− CML-like cells. Fig. S8 shows increased propensity to apoptosis of BCL6^−/− CML-like cells. Fig. S9 shows colony formation assays for normal BCL6^+/+ and BCL6^−/− hematopoietic progenitor cells. Fig. S10 summarizes results from telomere length measurements. Fig. S11 shows a meta-analysis of gene expression values of FoxO3a and BCL6. Fig. S12 shows recruitment of BCL6 to cell cycle regulators and the effects of BCL6-activation on cell cycle progression.

We would like to thank Riccardo Dalla-Favera for sharing BCL6^−/− mice generated in his laboratory and wild-type controls with us, Lothar Hennighausen for Stat5^fl/fl mice, David A. Fruman for sharing his FoxO3A reagents with us, and Arthur L. Shaffer and Louis M. Staudt for sharing their inducible BCL6 constructs. We would like to thank Michael R. Lieber and Michael Kahn (Los Angeles, CA) for critical discussions and for CML leukapheresis samples.

This work is supported by grants from the National Institutes of Health/NCI through R01CA137060 (to M. Müschen), R01CA139032 (to M. Müschen), R01CA157644 (to M. Müschen) and R21CA152497 (to M. Müschen), Translational Research Program grants from the Leukemia and Lymphoma Society (grants 6132-09 and 6097-10), a Leukemia and Lymphoma Society SCOR (grant 7005-11; B.J. Druker), the William Laurence and Blanche Hughes Foundation and a Stand Up To Cancer-American Association for Cancer Research Innovative Research Grant (IRG00909, to M. Müschen), and the California Institute for Regenerative Medicine (CIRM; TR2-01816 to M. Müschen). A.M. Melnick and M. Müschen are Scholars of the Leukemia and Lymphoma Society.

The authors have no conflicting financial interests.