Normal ABL1 is a tumor suppressor and therapeutic target in human and mouse leukemias expressing oncogenic ABL1 kinases

The ABL1 protein is a ubiquitously expressed nonreceptor tyrosine kinase markedly influenced by subcellular localization and posttranslational modifications.^1-3  Cytoplasmic expression of ABL1 leads to increased cell proliferation and survival. In response to genotoxic stress, ABL1 is translocated into the nucleus and/or mitochondria where its activity contributes to modulation of DNA repair, induction of apoptosis/necrosis, and inhibition of cell growth.

Normal ABL1 kinase activity is essential for B- and T-cell development, but expendable in hematopoietic stem cells (HSCs) and the myeloid compartment.^4-6  Constitutively activated oncogenic mutants of the ABL1 tyrosine kinase play a central role in the pathogenesis of acute and chronic leukemias. Activation usually occurs as a consequence of chromosomal translocations (BCR-ABL1, TEL-ABL1, and others) or episomal amplification (NUP214-ABL1).¹ 

The BCR-ABL1 fusion oncogene, the product of t(9;22)(q34;q11) is found in all patients with chronic myeloid leukemia (CML), in ∼25% of pre-B acute lymphocytic leukemia (ALL) and occasionally in de novo acute myeloid leukemia (AML).⁷  BCR-ABL1 kinase is leukemogenic only when expressed in an HSC with self-renewal capacity, thereby transforming it to a leukemia stem cell (LSC).⁸  In CML–chronic phase (CML-CP), LSCs are capable of generating large numbers of leukemia early progenitor cells (LPCs): leukemia common myeloid (LCMPs) and leukemia granulocyte/macrophage (LGMPs), which cannot self-renew and eventually differentiate to mature cells. Thus, CML-CP is a stem cell–derived but progenitor-driven disease.⁸  Transition of a relatively benign CML-CP to the aggressive and fatal blast phase (CML–blast phase [CML-BP]) is associated with expansion of LSCs, enhanced proliferation, arrested differentiation, drug resistance, and accumulation of additional genetic and epigenetic aberrations.^9,10 

NUP214-ABL1 fusion is generated by circularization of the 500-kb genomic region from ABL1 to NUP214 and subsequent extrachromosomal (episomal) amplification.¹¹  The NUP214-ABL1 gene is found in ∼4% of all cases of adult ALL. Other ABL1 fusion genes have been described but are uncommon. For example, the ETV6(TEL)-ABL1 fusion gene is the product of a t(9;12)(q34;p13) and is found occasionally in patients with acute leukemias or myeloproliferative disorders. EML1, ZMIZ1, and RCSD1 were identified as ABL1 partners in ALLs.¹ 

Leukemias expressing oncogenic forms of the ABL1 kinase usually contain the nonmutated allele encoding normal ABL1 kinase which may play an important role in pathogenesis of disease and/or in response to treatment, given its prominent role in regulation of cell motility, adhesion, autophagy, response to DNA damage, apoptosis, and proliferation.^1-3  This possibility is supported by previous observations that loss of normal ABL1 expression resulting from interstitial deletion in the normal chromosome 9 [del(9q34)] and/or transcriptional silencing of the alternative ABL1 promoter within BCR-ABL1 translocation occurs during progression of CML-CP to CML-BP.^12,13  Of note, in the absence of ABL1, BCR-ABL1 cells displayed reduced sensitivity to tyrosine kinase inhibitors (TKIs) such as imatinib.¹⁴  Therefore, we hypothesized that normal ABL1 is a tumor suppressor in CML-CP and therapeutic target in leukemias induced by oncogenic forms of ABL1 kinase.

BCR-ABL1–positive Abl1^−/− and Abl1^+/+ bone marrow cells (BMCs) expressing YFP-ABL1 fusion protein or yellow fluorescent protein (YFP) only were obtained and maintained as described in supplemental Methods (see supplemental Data available at the Blood Web site).

Green fluorescent protein (GFP)-positive or GFP/YFP-positive cells were injected into the tail vein of sublethally irradiated NOD/SCID mice. Animals were killed when first signs of disease were apparent and leukemia development was confirmed at necropsy. These studies were approved by the Temple University institutional animal care and use committee.

LSCs and LPCs were identified as described before¹⁵  and detailed in supplemental Methods.

Freshly transfected Lin⁻c-Kit⁺Sca-1⁺ BCR-ABL1 Abl1^+/+ and BCR-ABL1 Abl1^−/− cells were cultured for 5 weeks in vitro and simultaneously plated in MethoCult H4230 (StemCell Technologies, Vancouver, BC, Canada) in absence of growth factors. Colonies were scored after 5 to 7 days, and replated in fresh Methocult and scored again after 5 to 7 days. Three rounds of serial replating (representing 5 weeks in culture) were performed. Five-week-old tissue-cultured BCR-ABL1 Abl1^+/+ and BCR-ABL1 Abl1^−/− cells were also plated in Methocult. Colonies were scored after 5 to 7 days.

A mixture of GFP-positive BCR-ABL1 Abl1^−/− and GFP/YFP-positive BCR-ABL1 Abl1^−/− cells restored with YFP-ABL1 was maintained in Iscove modified Dulbecco medium (IMDM) supplemented with fetal bovine serum (FBS), stem cell factor (SCF), and interleukin-3 (IL-3) and also simultaneously injected into the tail vein of NOD/SCID mice. After 5 weeks, the in vitro and in vivo cell mixtures were analyzed by flow cytometry to determine the percentage of GFP and GFP/YFP double-positive cells.

Tissue sections of spleen from leukemia-sick mice were fixed in 10% formalin, paraffin embedded, cut into 0.4-mm sections, transferred to glass slides, and stained with hematoxylin and eosin. Cytospin preparations of cells cultured for 7 days in the presence of pretested threshold concentrations of IL-3 + SCF or granulocyte colony-stimulating factor (G-CSF; 10 ng/mL) were stained with Wright-Giemsa.

Indicated concentrations of cisplatin (Platinol-AQ; Bristol-Myers Squibb Co, Princeton, NJ), mitomycin C (Sigma-Aldrich, St. Louis, MO), etoposide (Bedford Laboratories, Bedford, OH), methylnitronitrosoguanidine (MNNG; Sigma-Aldrich), and hydrogen peroxide (Sigma-Aldrich) were added to cells growing in MethoCult H4230 (10³/mL) supplemented with IL-3. Colonies were scored after 7 days. Results are represented as the percentage of colony-forming cells after drug treatment in comparison with the untreated control group.

Reactive oxygen species (ROS), 8-oxoguanine (8-oxoG), γ-H2AX nuclear foci, and chromosomal aberrations were detected as described before¹⁶  and detailed in supplemental Methods.

CML-CP cells from freshly diagnosed patients, B-ALL xenograft cells, Baf3-TEL-ABL1 cell lines, and murine BMCs expressing NUP214-ABL1 oncogene were treated with 5-[3-(4-fluorophenyl)-1-phenylpyrazol-4-yl]imidazolidine-2,4-dione (DPH; Sigma-Aldrich), imatinib and ponatinib (both from Selleck Chemicals) and evaluated as described in supplemental Methods. These studies were approved by the Temple University institutional review board.

Total, cytoplasmic, and nuclear cell lysates were obtained as described before¹⁷  and analyzed as detailed in supplemental Methods.

The Affymetrix Mouse gene 1.0ST array containing 28 815 probe sets (Affymetrix) was used to measure messenger RNA (mRNA) expression levels. Affymetrix arrays were processed and analyzed as described in supplemental Methods.

Results are presented as mean ± standard deviation (SD) and were analyzed by the 2-tailed paired Student t test and the Mann-Whitney rank-sum test when appropriate. Median survival time (MST) of the mice was calculated by Kaplan-Meier log-rank survival analysis. P < .05 was considered statistically significant.

To determine the role of ABL1 in CML, BCR-ABL1 kinase was expressed in Abl1^−/− and Abl1^+/+ murine BMCs (BCR-ABL1 Abl1^−/− and BCR-ABL1 Abl1^+/+ cells, respectively); in addition, ABL1 kinase expression was restored in BCR-ABL1 Abl1^−/− cells by ectopic expression of YFP-ABL1 kinase as described before.¹⁴  Although 100% of NOD/SCID mice injected IV with BCR-ABL1 Abl1^−/− cells succumbed to leukemia, only 25% of mice injected with BCR-ABL1 Abl1^+/+ cells developed lethal disease (MST = 78.8 ± 5.9 days and 140.5 ± 8.2 days, respectively; P < .001) (Figure 1A). When BCR-ABL1–transduced cells were cultured in limited growth factor conditions for 5 weeks to achieve growth factor independence (5-week-old cells), all (100%) mice inoculated with 10³ BCR-ABL1 Abl1^−/− cells and only 67% of those injected with BCR-ABL1 Abl1^+/+ cells developed deadly disease (Figure 1B; MST = 36.2 ± 2.4 days and 105.7 ± 15.3 days, respectively; P < .001). Even though injection of 10⁵ of 5-week-old BCR-ABL1 Abl1^−/− and BCR-ABL1 Abl1^+/+ cells resulted in a deadly leukemia in all mice, the disease was accelerated in the absence of ABL1 (Figure 1C; MST = 19.7 ± 0.9 and 57.4 ± 5.5, respectively; P < .001). Injection of 10³ or 10⁵ of BCR-ABL1 Abl1^−/− cells reconstituted with YFP-ABL1 [BCR-ABL1 Abl1^−/−(+)] induced deadly leukemias in 58% and 100% mice, respectively, with latency (MST = 102.8 ± 12.7 and 45.8 ± 5.1, respectively) similar to that in animals injected with corresponding numbers of BCR-ABL1 Abl1^+/+ counterparts (Figure 1B-C).

Because uncontrolled outgrowth of LSCs is associated with CML-BP progression and/or therapeutic resistance,¹⁸  we sought to determine whether ABL1 affects the accumulation of stem cells in the BCR-ABL1–transformed BMC population. We found no significant difference in the percentage of Lin⁻c-Kit⁺Sca-1⁺ HSCs in freshly harvested Abl1^−/− (1.8 ± 0.5) and Abl1^+/+ (1.9 ± 0.6) bone marrow. In concordance, the percentage of Lin⁻c-Kit⁺Sca-1⁺ in freshly transduced GFP⁺ BCR-ABL1 Abl1^−/− and GFP⁺ BCR-ABL1 Abl1^+/+ cells did not differ (Figure 2A 0 weeks). However, after 5 weeks of in vitro culture, the percentage of LSCs derived from GFP⁺ BCR-ABL1 Abl1^−/− cells was over threefold higher than those derived from GFP⁺ BCR-ABL1 Abl1^+/+ cells (Figure 2A 5 weeks). Next, we analyzed GFP⁺ BCR-ABL1–positive Lin⁻c-Kit⁺Sca-1⁺ LSCs cultured for 5 weeks to quantify quiescent (Lin⁻c-Kit⁺Sca-1⁺eFluor670^high), long-term (LT) (Lin⁻c-Kit⁺Sca-1⁺CD34⁻) and short-term (ST) (Lin⁻c-Kit⁺Sca-1⁺CD34⁺) LSC subpopulations.¹⁵  An approximate threefold reduction of the percentage of quiescent LSCs was accompanied by ninefold and >100-fold increase in LT-LSCs and ST-LSCs, respectively, derived from GFP⁺ BCR-ABL1 Abl1^−/− cells in comparison with GFP⁺ BCR-ABL1 Abl1^+/+ cells (Figure 2B).

GFP⁺ BMCs obtained from mice with advanced leukemia originally injected with either freshly transduced or 5-week-old GFP⁺ BCR-ABL1 Abl1^+/+ cells contained 0.4% ± 0.3% GFP⁺Lin⁻c-Kit⁺Sca-1⁺ LSCs, whereas those inoculated with GFP⁺ BCR-ABL1 Abl1^−/− counterparts accumulated 2.4% ± 2.0% and 68.7% ± 16.9% GFP⁺Lin⁻c-Kit⁺Sca-1⁺ LSCs demonstrating sixfold and 60-fold expansion, respectively (Figure 2C).

Transcriptome analysis by microarrays comparing BCR-ABL1 Abl1^−/− vs BCR-ABL1 Abl1^+/+ cells revealed differential expression of 33 genes potentially involved in regulating stem cell–like characteristics (Figure 2D). In general, genes positively regulating “stemness” including Cd34, Kit, Ly6a, Hoxb, FoxO3, Smad5, and Smad4 were upregulated and genes inhibiting “stemness” such as Ly6e, Fzd6, Eya2, and Slamf1 were downregulated in BCR-ABL1 Abl1^−/− vs BCR-ABL1 Abl1^+/+ cells. Moreover, Ingenuity Pathway Analysis (IPA) suggests that upregulation of Smad4/5 stimulatory activity and downregulation of Smad3 inhibitory activity downstream of the transforming growth factor β (TGFβ) superfamily signaling (supplemental Figure 1) combined with overexpression of the receptor tyrosine kinase Kit may contribute to abundant expansion of BCR-ABL1 Abl1^−/− Lin⁻c-Kit⁺Sca-1⁺ LSCs.¹⁹ 

Although BCR-ABL1 stimulates proliferation, ABL1 is known to both promote as well as inhibit cell proliferation depending on the cellular context.^1,20  Therefore, we examined whether the tumor-suppressive function of ABL1 is mediated through regulation of cell proliferation. Freshly transduced Lin⁻c-Kit⁺Sca-1⁺ BCR-ABL1 Abl1^+/+ cells displayed 1.5 times higher clonogenic potential than BCR-ABL1 Abl1^−/− counterparts (Figure 3A 0 weeks), in concordance with other reports indicating that cytoplasmic ABL1 promoted cell cycle progression in CML-CP cells.²¹  Conversely, 5-week-old cultured BCR-ABL1 Abl1^−/− cells formed ∼1.5 times more colonies when compared with BCR-ABL1 Abl1^+/+ and BCR-ABL1/YFP-ABL1 Abl1^−/− counterparts (Figure 3A 5 weeks). Overall, during 5 weeks of continuous culture, BCR-ABL1 Abl1^−/− cells generated ∼6 times more clonogenic cells than BCR-ABL1 Abl1^+/+ cells (Figure 3B).

To test directly whether loss of ABL1 provides a growth advantage, competition assays were performed in which 10⁵ freshly transduced cells consisting of 10% GFP-positive BCR-ABL1 Abl1^−/− cells and 90% GFP/YFP-double positive BCR-ABL1/YFP-ABL1 Abl1^+/+ cells were maintained for 5 weeks either in liquid culture or in NOD/SCID mice. Flow cytometry detected ∼95% of GFP-positive and 5% of GFP/YFP-positive cells in the mixtures obtained in vitro and in vivo, indicating that the absence of ABL1 provided ∼200-fold growth advantage for BCR-ABL1–transformed cells (Figure 3C bars). Overgrowth of BCR-ABL1 Abl1^−/− cells was confirmed by western blot analysis demonstrating undetectable YFP-ABL1 expression in cell mixtures harvested after 5 weeks (Figure 3C western blot insets).

Transcriptome analysis by microarrays was consistent with a growth advantage of BCR-ABL1 Abl1^−/− cells; 19 proproliferative genes (eg, Vegfc, Vegfb, Pdgfa, Lif, Hgf, Lyn, Rac2) were upregulated and 12 antiproliferative genes (eg, Ifitm3, Il1a, Il1b) were downregulated in these cells as compared with BCR-ABL1 Abl1^+/+ cells (Figure 3D). In addition, IPA suggests that enhanced signaling from phospholipase Cγ1 (PLCγ1) and upregulation of cytoplasmic tyrosine kinases such as Lyn, Fyn, and Syk may contribute to accelerated proliferation rate of BCR-ABL1 Abl1^−/− cells overexpressing receptor tyrosine kinase Kit (supplemental Figure 2).^22-25  Moreover, upregulation of Pdgfa, Vegfb, Vegfc, and Csf1 expression in BCR-ABL1 Abl1^−/− cells implicates autocrine stimulation of cell growth.

Because moderate expression levels of BCR-ABL1 in the presence of ABL1 promote differentiation in CML-CP and differentiation arrest in CML-BP is associated with enhanced BCR-ABL1 expression and downregulation of ABL1, we sought to determine the role of normal ABL1 in the maturation of the leukemic cells.^26,27  Although all moribund mice injected with BCR-ABL1 cells developed splenomegaly (supplemental Figure 3), hematoxylin/eosin staining of spleen sections obtained from mice injected with BCR-ABL1 Abl1^−/− cells displayed myeloblastic morphology (Figure 4A −/−). Conversely, spleens harvested from moribund animals injected with BCR-ABL1 Abl1^+/+ and BCR-ABL1/YFP-ABL1 Abl1^−/− cells revealed moderate to complete myeloid differentiation of leukemic cells ranging from promyelocytes to myelocytes, metamyelocytes, and mature neutrophils (Figure 4A +/+ and −/−(+), respectively).

In vitro, in the presence of IL-3 + SCF, cultures of either BCR-ABL1 Abl1^−/− or BCR-ABL1 Abl1^+/+ cells were composed almost exclusively of myeloid precursors having deep blue cytoplasm, but the BCR-ABL1 Abl1^+/+ population contained more mature myeloid cells displaying abundance of cytoplasmic granules including seemingly fully differentiated neutrophils with hypersegmented multilobed nuclei (Figure 4B SCF + IL-3). In the presence of G-CSF, cultures of BCR-ABL1 Abl1^+/+ cells were predominantly composed of mature neutrophils with very few myeloid precursors whereas cultures of BCR-ABL1 Abl1^−/− cells contained mostly myeloid precursors and few neutrophils (Figure 4B G-CSF). Reconstitution of ABL1 expression in BCR-ABL1/YFP-ABL1 Abl1^−/− leukemia cells facilitated accumulation of myeloid precursors and/or mature neutrophils in the presence of SCF + IL-3 and G-CSF, respectively (Figure 4B, −/−(+)). Flow cytometry detected <1% of Gr-1⁺Mac-1⁺ cells in BCR-ABL1 Abl1^−/− population whereas >7% and >14% of BCR-ABL1 Abl1^+/+ cells were Gr-1⁺Mac-1⁺ in the presence of SCF + IL-3 and G-CSF, respectively (Figure 4C).

Transcriptome microarray analysis supported a myeloid differentiation arrest profile of BCR-ABL1 Abl1^−/− cells when compared with BCR-ABL1 Abl1^+/+ cells; 7 genes promoting differentiation were inhibited (Klf1, Lmo2, Il3, Pml, Tesc, Vegfa, and Zfp36) whereas 5 genes associated with differentiation arrest were preferentially expressed (Bcl6, Lyn, Id2, Inpp5d, and Kit) (Figure 4D).

BCR-ABL1 kinase promotes cell survival after genotoxic stress whereas ABL1 facilitates apoptosis by exerting opposite effects on caspase 9–mediated activation of caspase 3.^28,29  To elucidate the impact of ABL1 loss in BCR-ABL1 cells, BCR-ABL1 Abl1^−/− and BCR-ABL1 Abl1^+/+ cells were exposed to genotoxic agents such as cisplatin, mitomycin C, etoposide, MNNG, or hydrogen peroxide. Results of the clonogenic assay demonstrated that ABL1 enhanced the toxicity of DNA-damaging agents (Figure 5A). ABL1 has been reported to modulate survival/apoptosis by regulation of p53 and p73.^30-33  Western blot followed by densitometry analysis show more abundant expression of p53 (11- to 16-fold), phospho-serine 15 p53 (1.5-fold to twofold), and p73 (fivefold), resulting in caspase 3 activation in cisplatin-treated BCR-ABL1 Abl1^+/+ when compared with BCR-ABL1 Abl1^−/− leukemic cells (Figure 5B). BCR-ABL1 Abl1^−/− cells reconstituted with YFP-ABL1 [BCR-ABL1 Abl1−/−(+)] display a similar pattern of expression of these proteins to BCR-ABL1 Abl1^+/+ cells.

The proapoptotic gene expression signature of BCR-ABL1 Abl1^+/+ cells was also detected by transcriptional microarrays. Several proapoptotic genes including Rnf144b, Ccnb1ip1, Casp4, Casp12, Eya2, and Gadd45g were downregulated and 3 antiapoptotic genes Nek6, Malt1, and Card11 were upregulated in BCR-ABL1–positive Abl1^−/− cells compared with BCR-ABL1 Abl1^+/+ cells (Figure 5C). In addition, IPA suggests that proapoptotic signaling involving JNK1-dependent phosphorylation of p53 and BCL2 is downregulated in BCR-ABL1–positive Abl1^−/− cells (supplemental Figure 4).^34,35 

BCR-ABL1–positive LSCs and LPCs contain high levels of ROS-induced oxidative DNA damage such as 8-oxoG and DNA double strand breaks (DSBs), which contribute to accumulation of chromosomal aberrations.^16,36  By contrast, ABL1 prevents genomic instability.³⁷  To evaluate the role of ABL1 in modulating ROS-induced oxidative DNA damage in BCR-ABL1 leukemia cells, endogenous ROS were measured with RedoxSensor Red CC-1 dye which detects superoxide anion (⋅O₂⁻) and hydrogen peroxide (H₂O₂) and with dihydroethidium (DHE) which detects ⋅O₂⁻. BCR-ABL1 Abl1^−/− cells contained 1.5 and 2.5 times less ROS detected by CC-1 and DHE, respectively, when compared with BCR-ABL1 Abl1^+/+ cells (Figure 6A). Transcriptional microarray analysis suggests that lower levels of H₂O₂ and ⋅O₂⁻ in BCR-ABL1 Abl1^−/− cells were probably caused by 15-fold upregulation of glutathione peroxidase 3 (Gpx3) responsible for detoxification of H₂O₂ and 1.7-fold downregulation of mitochondrial electron transfer flavoprotein-ubiquinone oxidoreductase (Etfdh) which facilitates production of superoxide.^38,39  As predicted by previous reports, decrease in ROS levels in BCR-ABL1 Abl1^−/− cells was accompanied by twofold lower oxidative DNA damage measured by 8-oxoG as well as threefold decrease in the number of DSBs quantitated by γ-H2AX when compared with BCR-ABL1 Abl1^+/+ cells (Figure 6B-C).^16,40 

These unexpected findings prompted us to examine whether ABL1 regulated the accumulation of chromosomal aberrations in BCR-ABL1 leukemia cells. BCR-ABL1 Abl1^−/− and BCR-ABL1 Abl1^+/+ cells were cultured for 10 weeks followed by spectral karyotype analysis (SKY) to detect acquired chromosomal aberrations. Results from metaphase spreads revealed a nearly twofold increase in chromosome numbers and ∼2.5-fold higher number of chromosomal aberrations (whole chromosome and intrachromosomal gains/losses, as well as nonrecurrent chromosomal translocations) in BCR-ABL1 Abl1^−/− cells compared with BCR-ABL1 Abl1^+/+ cells (Figure 6D-F), consistent with overall high levels of genomic instability in the former cells.

The absence of ABL1 did not affect cell cycle distribution of BCR-ABL1 cells in response to DNA damage (supplemental Figure 5), but transcriptional microarray analysis revealed that the presence of ABL1 in BCR-ABL1 cells is associated with expression of numerous genes whose products regulate DNA damage response and mitotic spindle assembly checkpoint (Figure 6G-H). The majority of genes downregulated in BCR-ABL1 Abl1^−/− cells in comparison with BCR-ABL1 Abl1^+/+ cells are involved in promotion of DNA repair, for example, Ercc5, Fancd2, Polg, Exo1, Nhej1, Rif1, Rad18, Mre11, Msh6, Xrcc6, Top2b, and Atr. IPA suggested that simultaneous downregulation of Fancd2, and Xrcc6 and Nhej1 (involved in faithful homologous recombination and relatively faithful classical nonhomologous end-joining [C-NHEJ], respectively) may favor highly unfaithful PARP1-mediated alternative NHEJ resulting in facilitated accumulation of additional chromosomal aberrations in BCR-ABL1 Abl1^−/− cells (supplemental Figure 6).⁴¹  Genes involved in kinetochore/spindle/centrosome regulation (Pcid, Csnk2a2, Zwint, Plk2, Nek1) and sister chromatid segregation (Smc1a) are downregulated in BCR-ABL1 Abl1^−/− cells compared with BCR-ABL1 Abl1^+/+ cells, which may be responsible for abundant aneuploidy in the former cells.

Inhibition of intracellular ABL1 kinase may require threefold to 10-fold higher concentrations of imatinib than required to effectively inhibit BCR-ABL1 kinase⁴²  (supplemental Figure 7). For example, 1 to 2 µM imatinib, the concentration typically achievable in steady state in patients,^43,44  inhibits cytoplasmic BCR-ABL1, but it is suboptimal for inhibition of nuclear and mitochondrial ABL1.^45,46  Because ABL1 kinase enhanced the sensitivity of CML cells to imatinib,¹⁴  we hypothesized that stimulation of residual ABL1 activity may enhance the effect of TKIs in CML-CP treatment.

ABL1 allosteric activator DPH has been reported to stimulate ABL1 kinase by displacing the myristate moiety from the myristoyl-binding site, thereby preventing the autoinhibitory conformation of ABL1.⁴⁷  To determine whether DPH was able to induce and/or maintain the activity of ABL1 kinase under imatinib treatment, ABL1 was overexpressed in Phoenix cells followed by treatment with 10 μM DPH, 1 μM imatinib, or a combination of the 2 agents. DPH induced approximately eightfold and fourfold increase of phospho-Y245-ABL1, indicative of ABL1-kinase activation, in nuclear and cytoplasmic fractions, respectively (Figure 7A). Moreover, ∼50% to 70% of DPH-mediated activation of ABL1 kinase persisted in the presence of imatinib.

To determine whether ABL1 kinase could be a therapeutic target, we examined the effect of imatinib and/or DPH on BCR-ABL1 Abl1^+/+ and BCR-ABL1 Abl1^−/− cells. We found that treatment with 0.125 µM imatinib in combination with 10 µM DPH for 3 consecutive days reduced the number of BCR-ABL1 Abl1^+/+ leukemia cells when compared with single agents (Figure 7B). In contrast, BCR-ABL1 Abl1^−/− cells were not sensitive to the combination treatment thereby confirming the ABL1-specific effect of DPH. In addition, the combination of 1 µM imatinib and 10 µM DPH reduced clonogenic growth of CML-CP Lin⁻CD34⁺ cells by more than twofold in comparison with either agent alone, at the same time normal counterparts were not affected by the treatments (Figure 7C). Moreover, BCR-ABL1–positive B-ALL primary human xenograft cells and those harboring imatinib-resistant BCR-ABL1 (T315I) mutation were extremely sensitive (>10-fold) to the combination of imatinib (1 µM) + DPH (10 µM) and ponatinib (12.5 nM) + DPH, respectively, when compared with single-agent treated cells (Figure 7D-E).

Leukemia cells expressing TEL-ABL1 and NUP214-ABL1 oncogenic kinases were also more sensitive (more than twofold) to the combination of imatinib + DPH (Figure 7F-G).

We showed here that ABL1 exerts a tumor suppressor function in CML-CP by restricting the expansion of LSCs, inhibiting clonogenic activity of LSCs/LPCs, promoting myeloid differentiation and apoptosis, and reducing genomic instability. In concordance, loss of ABL1 facilitated development of highly malignant CML-BP–like disease in mice inoculated with BCR-ABL1 Abl1^−/− BMCs in contrast to CML-CP–like disease that arose from BCR-ABL1–transformed Abl1^+/+ cells.

The tumor suppressor function of ABL1 contrasting the oncogenic function of BCR-ABL1 is supported by reports that BCR-ABL1 and ABL1 can exert opposite effects on a variety of cellular functions.⁴⁸  We have shown that BCR-ABL1 kinase, which localizes exclusively to the cytoplasm,⁴⁹  protects leukemic cells from genotoxic stress, and facilitates genomic instability by promoting unfaithful DNA repair and deregulation of mitotic spindle.^50-55  Conversely, ABL1 which is targeted to the nucleus upon DNA damage can inhibit or stimulate DNA repair and stabilize mitotic spindle.^37,56-61  BCR-ABL1 inhibits apoptosis,⁶²  conversely nuclear and mitochondrial ABL1 kinase may facilitate apoptosis or necrosis.^28,31-33,63  Activation of the nuclear ABL1 inhibits cell proliferation but cytoplasmic BCR-ABL1 stimulates cell cycle progression.^64,65  These striking differences between ABL1 and BCR-ABL1 functions may explain the dramatic changes in proliferation, apoptosis, and genomic instability of BCR-ABL1 leukemic cells upon the loss of ABL1.

Remarkably, even though ABL1 did not affect myeloid differentiation of HSCs,^5,6  and BCR-ABL1 seemed to promote myeloid differentiation,²⁶  loss of ABL1 in BCR-ABL1 leukemic cells resulted in dramatic arrest of myeloid differentiation. Similarly, although ABL1 does not regulate the number of HSCs, BCR-ABL1 promotes rather mild expansion of LSCs in CML-CP^4,66  (and present study); nonetheless, rapid expansion of LSCs was observed in the absence of ABL1. It is plausible to speculate that a combined effect of enhanced proliferation and differentiation arrest resulted in rapid expansion of BCR-ABL1 Abl1^−/− LSCs.

The opposite effects of BCR-ABL1 and ABL1 may result from differences in their substrate preferences; in addition, their intracellular substrate repertoire may differ due to modification of kinase-substrate interactions and/or intracellular localization.^67,68  Accordingly, the exclusively cytoplasmic localization of BCR-ABL1 may allow it to escape the modulatory effect of intranuclear proteins such as retinoblastoma (Rb) tumor suppressor which negatively regulates ABL1.⁶⁹  Moreover, BCR-ABL1 demonstrated more abundant binding and modification of actin microfilaments than ABL1, which plays an important role in leukemogenesis.⁷⁰ 

Considering their opposite, pro- and antioncogenic properties, the BCR-ABL1-to-ABL1 ratio may play a key role in shaping the ultimate malignant features of the BCR-ABL1–transformed cells. It has been reported that CML-CP to CML-BP progression is associated with hypermethylation of the ABL1 promoter embedded in the translocation, which may be accompanied by clonal loss of expression of ABL1 transcribed from nontranslocated chromosome 9.^13,14,71-73  That would tip the balance toward BCR-ABL1 oncogene in individual clonogenic cells, which in conjunction with decreased sensitivity to imatinib¹⁴  might promote malignant progression. Moreover, patients undergoing imatinib treatment with yet undetected clones carrying BCR-ABL1 kinase mutations may be at greater risk of clonal evolution, eventually leading to CML-BP because even partial inhibition of ABL1 kinase (especially in patients treated with higher doses of imatinib) in the presence of intact BCR-ABL1 kinase mutant (eg, T315I) would promote chromosomal instability (supplemental Figure 8).

The fact that ABL1 kinase not only opposes BCR-ABL1 (present study) but also sensitizes CML cells to TKIs¹⁴  suggests that ABL1 activity could be explored therapeutically. The standard clinical dose of imatinib (400 mg per day) corresponds to trough plasma levels of ∼1 to 2 μM which effectively inhibited BCR-ABL1,^43,44  whereas ABL1 kinase required 3 to 10 μM of the drug to suppress its activity^45,46  (and this work). The tumor suppressor function of the intact ABL1 creates therapeutic opportunity for a combination of TKI and ABL1 activator. We showed here that allosteric activation of normal ABL1 kinase activity by DPH “hypersensitized” leukemic cells expressing oncogenic variants of ABL1 to TKIs in vitro. The beneficial antileukemia effect of imatinib + DPH was detected only in leukemia cells expressing ABL1, which supports the notion that DPH exerted its activity by activating ABL1, but not by displacing HSP90 from BCR-ABL1.⁷⁴ 

Optimization of DPH or other ABL1 activators may generate new drugs useful for development of novel therapeutic modalities against tumors expressing oncogenic mutants of ABL1.^75,76  For example, combination of TKI and allosteric ABL1 activator may enhance the therapeutic effect of TKIs in a cohort of CML-CP “poor responders” (no major cytogenetic response in 12 months, BCR-ABL1 transcript levels >10% after 3 months) and also in a majority of more aggressive BCR-ABL1–positive B-ALLs, which display unfavorable prognosis on TKIs.^77,78 

In summary, we showed here that the proto-oncogene ABL1 acts as a tumor suppressor in CML-CP cells expressing its oncogenic variant, BCR-ABL1. Moreover, analysis of the CGAP Mitelman database revealed that actual del(9q34) or deletions encompassing this region resulting in loss of ABL1 have been detected in other hematologic malignancies (see supplemental Table 1), thus suggesting that ABL1 may play a tumor suppressor role not only in CML. On the other hand, normal ABL1 kinase could be targeted to enhance the therapeutic effect of TKIs in leukemias driven by BCR-ABL1 and other oncogenic forms of ABL1 kinase. It is possible that the enhanced cytotoxic effect of TKI in the context of activation of ABL1 may not be limited to tumors expressing oncogenic forms of ABL1 but it could be a more general paradigm as supported by our findings of a similar effect in leukemia cells harboring the FLT3-ITD mutation when treated with the FLT3 inhibitor AC220 in combination with DPH (supplemental Figure 8).

The authors thank Dr John Tobias from the Molecular Profiling Core of the University of Pennsylvania (Philadelphia, PA) for the assistance with microarray analyses.

This work was supported by National Institutes of Health/National Cancer Institute CA134458 (T. Skorski), HARMONIA grant 2014/14/M/NZ5/00441 from the Polish National Science Center (T. Stoklosa), and by the Austrian Science Fund (FWF) F4704-B20 (P.V.).

Contribution: Y.D. performed in vitro and ex vivo experiments for stemness, proliferation, differentiation, DNA damage response, DPH treatment, and microarray analysis, and wrote the first draft of the manuscript; M.K. performed in vitro and in vivo experiments for proliferation and in vitro experiments for stemness, myeloid differentiation, apoptosis, and DNA damage response; G.H. performed in vivo experiments; M.N.-S. assisted with experiments using primary cells; K.K., D.R., M.A.W., and B.G. ran histopathological analyses and differentiation tests; E.B.​-G. assisted with flow cytometry data analysis; C.R. performed the SKY assay; S.C.-R., P.V., T. Stoklosa, and M.M. provided patient samples and primary leukemia xenografts, and performed CGAP Mitelman database analysis; O.H. provided NUP214-ABL1–transformed cells; H.v.d.K. performed the experiments in supplemental Figure 1; and T. Skorski conceived the idea, supervised the project, and wrote the final version of the manuscript.

Conflict-of-interest disclosure: P.V. served as a consultant and/or obtained honoraria and research funding from Novartis, Ariad, Bristol-Myers Squibb, Pfizer, and Celgene. The remaining authors declare no competing financial interests.

Correspondence: Tomasz Skorski, Department of Microbiology and Immunology and Fels Institute for Cancer Research and Molecular Biology, School of Medicine, Temple University, 3400 N. Broad St, MRB 548, Philadelphia, PA 19140; e-mail: tskorski@temple.edu.