BCR/ABL oncogenic kinase promotes unfaithful repair of the reactive oxygen species–dependent DNA double-strand breaks

The bcr/abl chimeric gene is derived from relocation of part of the c-abl gene from chromosome 9 to part of the bcr gene locus on chromosome 22 (t(9;22), Philadelphia chromosome [Ph]) and is present in most patients with chronic myelogenous leukemia (CML) and in a cohort of patients with acute lymphocytic leukemia (ALL).^1,2  BCR/ABL exhibits 2 complementary roles in cancer: it stimulates signaling pathways that render leukemia cells independent of their environment, and it modulates the response to DNA damage causing drug resistance.^3,4  In contrast to normal cells, BCR/ABL-positive cells seem to be better equipped to survive genotoxic damage because of their enhanced ability to repair DNA lesions, prolonged activation of the G₂/M checkpoint to provide more time for repair, and inhibited proapoptotic mechanisms.⁵  Clinical observations and experimental findings have shown that BCR/ABL also stimulates genomic instability, leading to mutations and chromosomal abnormalities.^6-12  The accumulation of genetic errors is believed to be responsible for the transition from a relatively benign CML chronic phase (CML-CP) to the aggressive blast crisis phase (CML-BC). Aberrations in pathways regulating the DNA damage response in BCR/ABL-positive leukemia cells may contribute to this phenomenon.

DNA damage can directly result from genotoxic treatment or may simply occur as a consequence of genome duplication infidelity or of genotoxic effects of compounds such as reactive oxygen species (ROSs). ROSs are generated as a normal byproduct of normal oxidative metabolism in eukaryotic cells, and they can cause damage to all molecules, including DNA.¹³  Single-strand oxidative DNA damage, including species such as 8-oxoguanine (8-oxoG), has been documented for many years.¹⁴  If not repaired, the damage can induce a number of deleterious effects, including mutations¹⁵  and double-strand breaks (DSBs),¹⁶  leading to elevated cancer risk.¹⁷ 

BCR/ABL kinase-stimulated ROSs¹⁸  may exert chronic genotoxic stress, causing DNA damage in leukemia cells. Given that mechanisms necessary for the repair of DNA lesions might be altered by BCR/ABL,^10,19-22  the probability of accumulating DNA errors seems to be much higher in BCR/ABL-positive cells than in nontransformed cells because more DNA damage occurs, and, though overall repair capability is enhanced, the fidelity of repair mechanisms is compromised. A mutator phenotype may be essential for tumor cells to grow in various organs under diverse conditions and to resist antitumor treatments.

This work shows that BCR/ABL elevates the level of ROSs, resulting in numerous DSBs during genome duplication and division (S and G₂/M phases). Unfortunately, the oncogene promotes unfaithful mechanisms of DSB repair. We hypothesize that elevated levels of DSBs, combined with unfaithful repair mechanisms, may contribute to genomic instability and malignant progression of Ph-positive leukemias.

The murine growth factor-dependent myeloid cell line 32Dcl3 and BCR/ABL- or BCR/ABL[K1172R]-expressing clones²⁰  have been maintained in the presence of pretested optimal concentrations of interleukin-3 (IL-3) required to maintain their continuous proliferation. Bone marrow mononuclear cells from C57Bl/6 mice (mBMCs) (The Jackson Laboratory, Bar Harbor, ME) were infected with breakpoint-cluster region/Abelson leukemia-internal ribosomal entry site-green fluorescence protein (BCR/ABL-IRES-GFP) or IRES-GFP retroviral particles, as described.²⁰  GFP-positive cells obtained after sorting were cultured for 72 hours in the presence of pretested concentrations of IL-3 and were used for experiments. Bone marrow cells from patients with CML (2 with CML chronic phase and 2 with CML blast crisis) and healthy donors (hBMCs), described before,²³  were obtained after informed consent. Mononuclear cells were cultured for 72 hours in the presence of pretested recombinant human IL-3 and stem cell factor (SCF), and CD34⁺ cells were isolated.²⁴  Draa-40²⁵  and BCR/ABLDraa-40 cells were cultured in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS).

Cells were treated with the following compounds: 2 μM imatinib mesylate (STI571; Novartis Pharma AG, Basel, Switzerland), 0.2 μM antioxidant pyrrolidine dithiocarbamate (PDTC), 100 μM nitrone spin traps N-tert-butyl-a-phenylnitrone (PBN) and 5,5-dimethyl-1-pyrroline-N-oxide (DMPO), and 2 μM wortmannin and 5 mM caffeine (Sigma-Aldrich, St Louis, MO). Imatinib mesylate, PDTC, PBN, and DMPO were added for 48 hours; wortmannin and caffeine were added for 3 hours.

Comet assay was performed under alkaline conditions, as described,²⁶  with modifications.²⁷  Comet tail moment was analyzed in 50 images randomly selected from each sample, in duplicate experiments (total, 100 images/sample); it positively correlates with the level of DNA breakage and alkali-labile sites.²⁸  The value of tail moment in particular samples was taken as an index of DNA damage. Because our measurement system was not calibrated, tail moment was presented in arbitrary units. Results represent mean ± SEM. Data were analyzed using the Statistica (StatSoft, Tulsa, OK) statistical package. For enzyme treatment, cells were drained in agarose and covered with an enzyme buffer (control) or with the enzyme (1 μg/mL endonuclease III [EndoIII] or formamidopyrimidine-DNA glycosylase [(Fpg]) in buffer, incubated for 30 minutes at 37°C, as described,²⁹  and the comet tail moment was analyzed as described here. Results obtained with buffer only were subtracted from these, obtained with an enzyme (enzyme treatment usually increased the detection of DNA damage by approximately 2-fold).

Genomic DNA was isolated from 3 to 5 × 10⁶ cells using the DNeasy Tissue Kit (Qiagen, Valencia, CA), and 2 to 3 μg was run (30 V for 15 minutes, followed by 70 V for few hours) in 2.5% agarose gel containing ethidium bromide and was then photographed.

Levels of intracellular ROSs were analyzed in cells growing in the presence of IL-3 using the redox-sensitive fluorochrome 2′,7′-dichlorofluorescin-diacetate.¹⁸ 

Draa-40 cells²⁵  and BCR/ABL-Draa-40 cells²⁰  have integrated 1 or 2 copies of the modified gene for GFP (DR-GFP, which contains an I-SceI site) as a recombination reporter and a fragment of the GFP gene (which contains a BcgI site) as a donor for homologous repair. A homologous recombination repair (HRR) event restores functional GFP expression (BcgI positive), which is readily detected by flow cytometry.

Nonhomologous end joining (NHEJ) was measured in cell-free extracts as described,³⁰  with modifications.³¹  Briefly, 10⁷ cells were washed 3 times with ice-cold phosphate-buffered saline (PBS), and cytoplasm lysis was performed on ice for 10 minutes in hypotonic buffer A (10 mM HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid], 1.5 mM MgCl₂, 10 mM KCl, pH 7.5) and in proteinase inhibitors (2 μg/mL aprotinin, 2 μg/mL leupeptin, 0.5 mM phenylmethylsulfonyl fluoride [PMSF], 0.5 mM dithiothreitol [DTT], 25 mM NaF, 0.2 mM NaVO₃). Cell pellets were spun down for 3 minutes in 6000g at 4°C, and cytoplasmic lysis was repeated. A pellet containing nuclei was resuspended in buffer B (20 mM HEPES, 25% glycerol, 500 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA [ethyleneglycotetraacetic acid], pH 7.5) and in proteinase inhibitors as in buffer A, and it was rapidly frozen and thawed 3 times (liquid nitrogen/37°C). Whole mixture was centrifuged at 30 000g for 30 minutes at 2°C. Supernatant was dialyzed overnight against buffer C (25 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10% glycerol) containing proteinase inhibitors as in buffer A. Aliquots of nuclear protein samples were stored in -70°C. NHEJ reactions were performed under the following conditions. A mixture containing 10 μg nuclear lysate, 1 mM adenosine triphosphate (ATP), 0.25 mM dNTPs, 25 mM Tris-acetate, pH 7.5, 100 mM potassium acetate, 10 mM magnesium acetate, and 1 mM DTT was preincubated in 37°C for 5 minutes. The substrate—200 ng linear plasmid pBluescript KS+ (digested XhoI + XbaI to generate noncompatible 5′ overhangs)—was added to the reaction mix and incubated for 1 hour at 37°C. Samples were then incubated with 1 μg proteinase K in 65°C for 30 minutes. Products of NHEJ reaction were resolved in 0.5% agarose gel containing 0.5 μg/mL ethidium bromide, scanned with Adobe Photoshop, and analyzed by ImageQuant TL (Amersham Bioscience, Piscataway, NJ).

Draa-40 parental and BCR/ABL-Draa-40 cells were transfected with I-SceI expression plasmid, and the frequency and fidelity of DSB repair was examined as described by Pierce et al,³²  with modifications. Briefly, 3 days after I-SceI transfection, a fragment of the DR-GFP cassette containing a putative DSB repair site was amplified by polymerase chain reaction (PCR) using DR-GFP cassette-specific primers: forward (A), CAGCCATTGCCTTTTATGGT; reverse (B), GCCTGAAGAACGAGATCAGC. Products were cloned into the TA cloning kit (Invitrogen, Carlsbad, CA) plasmid and were transformed into the One Shot TOP10. Bacterial colonies containing BcgI site (HRR product) or I-SceI site (replication product) or not containing either site (NHEJ product) were identified by Southern dot blotting using the BcgI- and I-SceI specific probes GGTGGCATCGCCCTCGCC and GGTATTACCCTGTTATCCCTAGCCGGA, respectively (hybridization conditions allowed detection of a sequence containing one mismatched base). The presence of the NHEJ product in the BcgI/I-SceI double-negative colonies was confirmed by PCR. Repair products were amplified from bacteria by PCR using the forward (C) AGGGCGGGGTTCGGCTTCTGG and reverse (D) CCTTCGGGCATGGCGGACTTGA primers to amplify the HRR products³²  and commercially available T3 and T7 primers to amplify the NHEJ products (primers spanning larger fragments of DR-GFP cassette were used here to reduce the chance of omitting more extensive deletions), and sequenced. HRR mutation frequency was calculated as total number of mutated nucleotides/total number of sequenced nucleotides. Identical mutations detected in BcgI-positive sequences and in I-SceI-positive sequences (20 sequences/group analyzed) were subtracted from the calculation to exclude replication errors. An average gain/loss of the bases in the double-negative BcgI-/I-SceI-NHEJ sequences was determined by dividing the sum of acquired/deleted bases by the number of sequences. Twenty repair products (HRR and NHEJ)/experimental group were analyzed.

Nuclear localization of the indicated proteins was detected by immunofluorescence, as previously described.¹⁰  Briefly, cytospins from unsynchronized cells were fixed in PBS with 0.06% Triton X-100 and 4% formaldehyde, washed in PBS and 0.06% Triton X-100, and blocked in washing buffer supplemented with 1% bovine serum albumin (BSA). To perform cell cycle-specific analysis, cells were washed with PBS, spun down, gradually resuspended in ice-cold fixing buffer (PBS with 0.06% Triton X-100, 4% formaldehyde, and 0.04% glutaraldehyde), and incubated on ice for 30 minutes. Fixed cells were washed twice with PBS, resuspended in 1 mL PBS with 100 μg RNase DNase-free (Roche, Mannheim, Germany), and incubated at room temperature for 15 minutes. DNA was stained with propidium iodide as described.⁵  G₀/G₁, S, and G₂/M cell populations were isolated by fluorescence-activated cell sorter (FACS) and cytospun. FACS was used to confirm the cell cycle purity of sorted cell populations. Cells were stained with first antibodies against γ-H2AX (Upstate Biotechnology, Lake Placid, NY) and RAD51 (PC130; Oncogene Research Products, Cambridge, MA) or Ku70 (AHP318, Serotec, Raleigh, NC). Secondary antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 568 were applied (Molecular Probes, Eugene, OR). Negative controls were performed without primary antibodies. DNA was counterstained with DAPI (4′,6′diamedino-2-phenylindole). Specific staining was visualized using an inverted Olympus IX70 fluorescence microscope equipped with 100 × UPlan Apo lens (numeric aperture 1.35), and a Cooke Sensicam QE camera (The Cooke Company, Auburn Hills, MI). At least 50 individual cells were analyzed per experimental group, as described.¹⁰  Images were acquired with Slidebook 3.0 (Intelligent Imaging Innovations, Denver, CO). All graphic adjustments were performed using Adobe Photoshop.

Genomic DNA was purified using the DNeasy Tissue Kit (Qiagen). The protocol for linker-ligation PCR (LL-PCR) to detect broken-ended, double-stranded DNA, described by Schlissel et al,³³  was followed with modifications. Briefly, the 2 oligomers BW-1 (GCGGTGACCCGGGAGATCTGAATTC) and BW-2 (GAATTCAGATC) were annealed to form a linker and were stored frozen. Purified DNA (2 μg) was incubated with Klenow polymerase and ligated to the linker. Ligated DNA (200 ng) was used in a 50 μL PCR assay containing 3 pmol linker reverse primer (BW-1) and DR-GFP cassette-specific forward primer (AGGGCGGGGTTCGGCTTCTGG) or Na⁺/K⁺ ATPase-specific forward primer (GCATTGACTTGGGCACTGAC) and 2 U Taq polymerase. After 2 rounds of PCR (2 × 20 cycles), the products were detected by Southern blot analysis using α-dCTP-labeled DR-GFP probe (NarI-HpaI fragment of the DR-GFP gene) and γ-ATP labeled Na⁺/K⁺ ATPase-specific probe (GGCCTACGAGCAAGCTGAG oligomer), respectively.

Comet assay indicated that the presence of BCR/ABL kinase increased the level of DNA damage (strand breaks and abasic sites) on average by approximately 2.5-fold. Inhibition of the kinase by 48-hour incubation with 2 μM imatinib mesylate³⁴  or ROSs by 48-hour incubation with the antioxidants (0.2 μM PDTC,¹⁸  100 μM PBN, and 100 μM DMPO³⁵ ) reduced the damage to the level detected in parental cells (Figure 1A). Imatinib mesylate and the antioxidants inhibited intracellular ROSs (data not shown). In agreement, gel electrophoresis also detected chromosomal fragmentation of the genomic DNA in BCR/ABL cells, but not in parental cells or in BCR/ABL cells treated with imatinib mesylate or PDTC (Figure 1B), implicating DNA breaks occurring specifically in a BCR/ABL kinase-dependent, ROS-mediated manner. To directly detect the oxidative damage of DNA, we used 2 enzymes, EndoIII and Fpg, which convert oxidative lesions into gaps detectable by the comet assay. Adding EndoIII or Fpg to the comet reaction significantly increased the DNA damage in BCR/ABL cells relative to parental cells; again, oxidative DNA damage levels were abrogated by the inhibition of BCR/ABL kinase with imatinib mesylate and the reduction of ROSs by PDTC (Figure 1C).

To characterize the DNA breaks in BCR/ABL cells, histone γ-H2AX was used to quantitate DSBs and to spatially localize them within the nuclei of cells. γ-H2AX seems to result from a specific ATM/ATR-dependent phosphorylation of histone H2AX on serine 139 in chromosome regions encompassing megabase lengths of DNA adjacent to DSBs.^36-38  Approximately 60% of the exponentially growing BCR/ABL-32Dcl3 cells contained γ-H2AX foci compared with 20% of parental cells (Figure 2A). To confirm that this effect occurs also in primary cells, mBMCs were infected with BCR/ABL retroviral particles. Approximately 61% of BCR/ABL-positive growth factor-independent mBMCs displayed γ-H2AX foci, in contrast to only 27% of normal counterparts (Table 1). Moreover, CML primary cell populations contained more γ-H2AX foci-positive cells than normal cells (17% and 2%, respectively; 4 patients with CML and 3 healthy donors were analyzed). The generally lower levels of γ-H2AX foci-positive healthy human cells, compared with their murine counterparts, may depend on the differences in basal oxidative damage levels; the former cells contain approximately 10 times fewer oxidized lesions than the latter ones.³⁹ 

The γ-H2AX foci were formed in response to DNA damage-dependent activation of ATM/ATR kinases because wortmannin and caffeine (inhibitors of these kinases⁴⁰ ) dramatically reduced the number of foci (Table 1). Inhibiting BCR/ABL kinase by imatinib mesylate and blocking of ROSs by PDTC diminished the percentage of γ-H2AX-positive BCR/ABL cells (Table 1). This result implicates the BCR/ABL kinase-dependent, ROS-mediated mechanism leading to the elevated levels of leukemia cells bearing DSBs. In addition to the increased percentage of BCR/ABL-positive cells displaying γ-H2AX foci, these cells contained at least 2-times more foci per cell than their nontransformed counterparts (Figure 2A). The percentages of γ-H2AX-positive cells containing 10 or less, and more than 10 foci per nucleus were 3% and 97%, respectively, in BCR/ABL-32Dcl3 cells, in contrast to 75% and 25%, respectively, in parental 32Dcl3 cells.

To determine whether γ-H2AX foci were associated with a particular cell cycle phase, exponentially growing 32Dcl3 parental cells and BCR/ABL-32Dcl3 cells were sorted on the G₀/G₁, S, and G₂/M populations according to DNA content. Analysis of cell cycle distribution of γ-H2AX-positive cells revealed that most cells were in S and G₂/M phase but not in G₀/G₁ phase; BCR/ABL did not change their percentage distribution (Figure 2B).

LL-PCR was used to detect DSBs to confirm the results obtained with the use of γ-H2AX foci as a marker of DSBs. We found 2 hotspots for DSBs in the DR-GFP gene in BCR/ABLDraa-40 cells. These DSBs were not detectable in parental cells or in BCR/ABL cells if ROSs were inhibited by PDTC (Figure 2C, left panel). In addition, DSBs in the α1 subunit of the Na⁺/K⁺ATPase gene were detected by LL-PCR in BCR/ABL-32Dcl3 cells but not in parental 32Dcl3 cells (Figure 2C, right panel). Again, PDTC inhibition of ROSs prevented detection of these DSBs in BCR/ABL cells. Combined analysis of the approximate lengths of LL-PCR products, position of the forward primers, and DR-GFP or Na⁺/K⁺ATPase gene sequences revealed multiple G/C-rich sequences at the predicted DSB hotspots (Figure 2C).

As expected, increased levels of ROS-dependent DSBs in the DR-GFP gene in BCR/ABL cells were associated with increased levels of spontaneous HRR, as measured by the detection of GFP-positive cells after 8 weeks of cell culture. HRR rate (estimated using Luria-Delbruck fluctuation analysis, with modifications⁴¹ ) was 30 × 10^-4 and 3.5 × 10^-4 for BCR/ABLDraa-40 cells and Draa-40 parental cells, respectively. BCR/ABLDraa-40 cells incubated in the presence of PDTC displayed a reduced HRR rate, 11 × 10^-4, compared with untreated BCR/ABL-Draa-40 cells.

Because most cells containing DSBs are in the S and G₂/M cell cycle phases, DSBs could be repaired by either of the 2 competing machineries, HRR or NHEJ.^42-44  The efficiency of HRR and NHEJ in BCR/ABL-positive and -negative cells was examined using the specific tests.

A copy of the DR-GFP cassette containing inactivated GFP gene, resulting from the introduction of the unique I-SceI restriction site containing 2 stop codons, and a truncated version of the gene were integrated into the genome of Draa-40 cells.²⁵  DSB is generated in the GFP gene upon transient transfection with I-SceI expression plasmid, which could be repaired by HRR using the truncated fragment of the gene as a template, thus producing a functional gene and, hence, GFP-positive cells. Draa-40 cells containing the DR-GFP cassette were transfected with expression plasmids containing BCR/ABL wild-type, BCR/ABL[K1172R] kinase-inactive mutant (BCR/ABL_[kin-]), or empty plasmid; the expression of BCR/ABL proteins was confirmed by Western blot analysis (not shown). These cells were then transfected with I-SceI expression plasmid (and the β-galactosidase plasmid to control the transfection efficiency), and HRR frequency was measured by scoring the percentage of GFP-positive cells. The presence of BCR/ABL active kinase, but not the kinase-dead mutant, caused an approximately 4-fold increase in the percentage of GFP-positive cells, implicating activation of the HRR mechanism (Figure 3A). pBluescript plasmid linearized by XhoI + XbaI digestion creating noncomplementary 5' overhangs was used as the substrate to assess the activity of NHEJ, which generates the products containing multimers of plasmid.^30,45  The substrate was added to cell lysates from 32Dcl3 parental, BCR/ABL wild-type, and BCR/ABL[K1172R] kinase-inactive (BCR/ABL_[kin-]) cells, and NHEJ products were analyzed by agarose gel electrophoresis. BCR/ABL kinase activity was responsible for more than a 5-fold increase of NHEJ activity; inactivation of the kinase by point mutation (K1172R) reduced NHEJ by 3-fold (Figure 3B).

HRR and NHEJ reaction sites in the nuclei could be potentially visualized by double-immunofluorescence detecting colocalization of γ-H2AX foci with RAD51 or Ku70, respectively.⁴⁶  As shown in Figure 3C, approximately 50% of DSBs marked as γ-H2AX foci colocalized with RAD51 in parental and BCR/ABL-positive 32Dcl3 cells; similarly, approximately 50% of the γ-H2AX foci colocalized with Ku70. This suggests that HRR and NHEJ are equally important for the repair of ROS-mediated DSBs in BCR/ABL cells.

Parental cells treated for 48 hours with 100 μM buthioninesulfoximine (BSO), which inhibits γ-glutamyl-cysteine synthetase, thus increasing ROSs by depleting glutathione pools,⁴⁷  displayed an enhancement of ROSs (data not shown) and an increased percentage (56 ± 11) of γ-H2AX-positive cells (compare with group P in Table 1). However, the elevation of ROS-mediated DSBs in parental cells was not associated with enhanced HRR (1.36% ± 0.3% of GFP-positive cells) and NHEJ (5.8% ± 1.8% of repair) compared with the untreated cells (Figure 3A-B, group P). Finally, approximately 10% to 20% of BSO-treated cells have been displaying an apoptotic signature (subdiploid amount of DNA) after 72 hours of incubation. These effects were reversed by the antioxidant PDTC (data not shown), implicating ROSs dependence.

Parental and BCR/ABL-Draa-40 cells were transfected with I-SceI expression plasmid to inflict a DSB in the reporter DR-GFP cassette. The primers spanning a fragment of the DR-GFP cassette containing the DSB site were used 72 hours later to amplify the repair products by PCR, which were then analyzed by Southern assay and sequenced to determine the frequency and fidelity of a repair mechanism (Figure 4A). Sequences with restored BcgI restriction site were considered HRR product, those without BcgI and I-SceI sites were identified as NHEJ products, and those with preserved I-SceI site are considered the replication products (DSB and repair did not occur).

Combined PCR-Southern blot analysis revealed that HRR and NHEJ consist of approximately 40% and 60% of DSB repair products, respectively, and that BCR/ABL does not significantly change this proportion (Figure 4A, % of repair events), though it enhances HRR and NHEJ activity (Figure 3A-B). HRR products obtained from parental cells were repaired faithfully (no mutations found), whereas the products from BCR/ABL-positive cells contained mutations (overall mutation rate, 6 × 10^-3) (Figure 4A, HRR branch). Analysis of the mutations revealed that almost 55% of mutations involved G/C → A/T transitions (40.4%) and G/C → T/A transversions (14.3%), 2 of the most common mutations resulting from oxidative DNA damage.¹⁵  These mutations do not cluster near the DSB site (Figure 4B). Typical deletions or additions (range, +3 to -59 bp) were detected in NHEJ products in parental cells (Figure 4A, NHEJ branch). NHEJ products in BCR/ABL cells contained only deletions (range, -2 to -533 bp); 35% of the products lost more than 100 bp. On average, the presence of BCR/ABL induced a statistically significant (P = .01) loss of DNA during NHEJ.

A mutator phenotype in cancer cells represents a major factor contributing to malignant progression of the disease.⁴⁸  In general, 2 aberrant functions may be attributed to genomic instability: enhanced DNA damage and compromised DNA repair mechanisms. DNA damage can occur as a result of the activity of endogenous compounds such as ROSs and exogenous factors such radiation or genotoxic compounds.⁴⁹  The DNA repair machinery may be deregulated by loss or gain of function.^50,51 

BCR/ABL cells treated with genotoxic agents displayed higher levels of DNA damage and aberrant DNA repair mechanisms.^{5,19,20,22,52-54}  These factors, combined with impaired apoptotic response and prolonged activation of the G₂/M checkpoint, can lead to genomic instability in BCR/ABL cells surviving the treatment.⁴  However, we show here that even untreated BCR/ABL cells contained elevated levels of DNA lesions (including DSBs) compared with nontransformed cells. This effect—in agreement with other findings showing constitutive DNA damage in CML cells⁵³ —was observed not only in cell lines but also in primary cells, such as CML cells and murine bone marrow cells transformed by BCR/ABL. A recent report by Dierov et al⁵⁵  failed to detect spontaneous DNA damage shortly after tetracycline-induced BCR/ABL expression, suggesting that the effect may require a more sustained presence of BCR/ABL.

We found that elevated levels of DSBs in BCR/ABL cells, when compared with levels of their normal counterparts, resulted from ROS-dependent oxidative DNA damage. Stimulation of parental hematopoietic cell lines by a growth factor⁵⁶  or a BCR/ABL kinase¹⁸  resulted in elevations of ROSs' levels compared with growth factor-starved parental cells. We cultured fully transformed growth factor-independent BCR/ABL cells and parental cells in the pretested optimal concentrations of IL-3 necessary to maintain continuous proliferation of the latter cells. In these conditions, ROSs were modestly (approximately 2-fold) enhanced in the former cells (data not shown). This effect may not be the sole factor contributing to the elevated oxidative damage in BCR/ABL cells. Degradation of the ROS-dependent oxidized nucleotides in the cytoplasm, their incorporation rate into DNA by polymerases, and repair efficiency of the oxidized bases incorporated into DNA may be modified in BCR/ABL cells.⁵⁷  In addition, BCR/ABL may provide necessary protection against apoptosis induced by these DNA lesions (eg, DSBs), allowing an accumulation of cells with damaged DNA.^58-60  This hypothesis is supported by our observation that elevations of ROS-induced γ-H2AX foci in parental cells treated with BSO were associated with apoptosis in some cells.

ROS can induce a variety of DNA lesions, such as oxidative base damage (eg, 8-oxoG) resulting from the incorporation of the oxidatively modified nucleotides (eg, 8-oxoGTP) into DNA by polymerases.⁶¹  If not repaired properly, these lesions may lead to mutations, topoisomerase I-oxidized DNA cleavage complexes, DSBs at replication forks, and chromosome breaks detected on metaphase spreads.^15,16,62,63  Usually, oxidized bases are repaired by base excision repair, which requires the activity of a DNA glycosylase and an AP endonuclease to generate gaps in single-stranded DNA.⁶⁴  Such interruptions, if encountered by a replication fork, can cause DSBs,⁶²  which can be detected experimentally as γ-H2AX foci.⁶⁵  We hypothesize that ROS-mediated oxidative DNA lesions in BCR/ABL cells produce numerous DNA repair intermediates, which do not induce the G₁/S checkpoint in BCR/ABL cells,^5,66  resulting in DSBs at the replication forks. This speculation is supported by the finding that BCR/ABL cells containing numerous γ-H2AX foci (marker of DSBs) were detected almost exclusively in S and G₂/M phase, but not in G₀/G₁ phase. This is in agreement with the previous observation that G₁ cell cycle phase arrest prevented ROS-induced DNA strand breaks.⁶⁷  We were able to detect a few cells with γ-H2AX foci in G₁ phase, in agreement with a report that ROS caused DSBs in serum-starved G₁-phase fibroblasts.⁶⁸ 

We suggest that ROS-dependent DSBs present in BCR/ABL cells occurred in the regions containing multiple 5- to 9-bp stretches of G/C. This supports findings that oxidative damage is predominantly detected in G/C-rich sequences.⁶⁹  Thus, it is intriguing to speculate that attempts to repair extensive clustered oxidative damage in BCR/ABL cells may generate DNA lesions that pose a serious obstacle for replication forks and result in DSBs. Lower levels of ROSs in normal cells may cause limited DNA damage, resulting in fewer DSBs, as suggested by the detection of fewer γ-H2AX foci. Interestingly, LL-PCR did not detect DSBs in parental cells, which may simply reflect the limited sensitivity of the assay or the formation of γ-H2AX foci on the lesions that are different from those of DSBs. γ-H2AX foci may also be detected on single-stranded DNA at arrested replication forks.⁷⁰  We speculate that less abundant oxidative damage in parental and PDTC-treated BCR/ABL cells may cause stalling of a replication fork, which could be resolved by fork regression followed by reverse branch migration, without generation of a DSB. On the other hand, extensive oxidative DNA damage in BCR/ABL-positive cells may poise the mechanisms of fork regression/resetting, more frequently resulting in DSBs. Suggested defects in the activation of the intra-S-phase checkpoint in BCR/ABL cells may also facilitate the generation of DSBs.⁵⁵ 

BCR/ABL cells have to develop protective mechanisms in response to ROS-dependent chronic genotoxic stress causing DSBs (eg, prevention of caspase-3 activation and stimulation of DSB repair).⁴  HRR and NHEJ represent 2 major mechanisms of DSB repair.⁷¹  We show here that most ROS-mediated DSBs occur in S and G₂/M cell cycle phases, when both mechanisms are active and are competing for a DSB substrate.^42-44  This work supports previous findings that BCR/ABL stimulates HRR²⁰  and NHEJ^52,53  reactions. Moreover, RAD51 and Ku70 colocalize with γ-H2AX foci (DSB marker) with the same relative frequency, suggesting that HRR and NHEJ have an equal opportunity to be used. Genetic analysis of the DSB repair products supports this hypothesis; approximately 40% and 60% of these products were repaired by HRR and NHEJ, respectively. The modest shift toward NHEJ frequency, in comparison with that suggested by colocalization studies, may be explained by the observation that repair of some DSBs may be initiated by homologous invasion and completed by an NHEJ mechanism.⁷²  In summary, we hypothesize that HRR and NHEJ play important roles in the repair of ROS-mediated DBSs in BCR/ABL cells.

We show that though they are enhanced, DSB repair mechanisms are not faithful in BCR/ABL cells. Mutations and large deletions were detected in HRR and NHEJ products, respectively. Because the mutations are scattered, not clustered near the DSBs induced by I-SceI or those detected by LL-PCR, the mechanisms responsible for ROS-dependent DSBs may be different from those causing HRR-mediated mutagenesis. We hypothesize that clustered oxidative damage in G/C-rich regions may cause DSBs, whereas DNA polymerase-mediated mispairing opposite an oxidized base during HRR-dependent DNA replication may be responsible for the mutagenic effect. In addition, error-prone DNA polymerases, such as polymerase β (polβ), the expression of which is elevated in BCR/ABL cells,⁹  may eventually replace other polymerases usually involved in DNA replication during HRR.⁷³  Interestingly, base misincorporations made during DSB repair in Saccharomyces cerevisiae were not substrates for the mismatch repair machinery.⁷⁴  Therefore, mismatched bases incorporated to the HRR sites might not be removed efficiently, causing mutations in the recombination products. The molecular explanation for extensive degradation of DSBs preceding NHEJ in BCR/ABL cells is unknown. Aberrantly regulated exonucleases such as ExoI, Mre11, and Artemis could be suggested as potential mediators of this reaction.⁷⁵ 

CML cells accumulate genetic abnormalities during the course of the disease.⁷⁶  The most frequently observed involve additional chromosomes (Ph, +8, +19); isochromosome i(17q) associated with the loss of p53; reciprocal translocations (3;21 and 7;11) associated with the expression of AML-1/Evi-1 and NUP98/HOXA9 fusion proteins, respectively; other translocations and inversions associated with AML/myelodysplasia (inv(3), t(15;17)); loss of heterozygosity (LOH) at 14q32; homozygous mutations/deletions of pRb and p16/ARF; and mutations in p53 and RAS. The origin of these effects is mostly unknown, but errors in the repair of ROS-dependent DSBs described here can lead to the latter phenomena, especially to these involving intrachromosomal deletions and point mutations. Therefore, we hypothesize that genetic instability leading to the malignant progression of Ph-positive leukemias is associated with unfaithful repair of ROS-mediated DSBs.

In addition to BCR/ABL, myeloid cells transformed with other forms of the oncogenic ABL kinase, TEL/ABL and v-ABL, display the symptoms of ROS-dependent genotoxic stress (data not shown). Thus, unfaithful repair of ROS-dependent DSBs may represent a more universal mechanism of gaining a mutator phenotype in tumors caused by the ABL kinases and possibly also in those expressing other fusion tyrosine kinases (FTKs; BCR/FGFR, TEL/ABL, TEL/JAK2, TEL/PDGFβR, NPM/ALK).⁷⁷  Moreover, a recent report by Heath and Cross⁷⁸  shows that hematopoietic cells transformed by another FTK, ZNF198/FGFR1, which is associated with a disease that closely resembles BCR/ABL-positive CML, stimulate signal transducer and activator of transcription 5 (STAT5) to induce RAD51 and Bcl-xL expression, in accordance with our finding regarding other FTKs.⁵  Thus, cells transformed by FTKs seem to be prepared to handle increased levels of ROS-dependent DSBs.