Targeting homologous recombination using imatinib results in enhanced tumor cell chemosensitivity and radiosensitivity

RAD51 is a key protein in the homologous recombination (HR) pathway of DNA double-strand break repair, and HR represents a novel target for cancer therapy. Because imatinib (Gleevec) has been reported to reduce RAD51 protein levels, we tested the clonogenic survival for RT112, H1299, PANC1, and PC3 tumor cell lines of varying p53 status and normal GM05757 normal fibroblasts after exposure to single agent imatinib (0–20 μmol/L; 0–72 hours). We also combined imatinib with DNA damaging agents that are toxic to RAD51-deficient cells, including ionizing radiation, gemcitabine, and mitomycin C. We observed decreased nuclear expression and chromatin binding of RAD51 protein following imatinib treatment. Imatinib also resulted in decreased error-free HR as determined by a flow cytometry–based integrated direct repeat-green fusion protein reporter system; this correlated to reduced RAD51 expression. Clonogenic survival experiments revealed increased cell kill for imatinib-treated cells in combination with ionizing radiation, gemcitabine, and mitomycin C, due in part to mitotic catastrophe. In experiments using imatinib and gemcitabine, tumor cell lines were sensitized to a greater extent than normal fibroblasts. This preservation of the therapeutic ratio was confirmed in vivo using PC3 xenograft growth delay and intestinal crypt cell clonogenic assays. HR inhibition may be an additional mechanism of action for the chemosensitization and radiosensitization of solid tumors with imatinib with preservation of the therapeutic ratio. [Mol Cancer Ther 2009;8(1):203–13]

DNA is a major target for many cancer therapeutics including cytotoxic chemotherapy and radiotherapy. These agents can induce DNA double-strand breaks (DNA-dsb) whereby unrepaired lethal lesions lead to apoptosis, mitotic catastrophe, terminal growth arrest, and finally, increased clonogenic tumor cell kill. Human DNA-dsbs are primarily repaired by the homologous recombination (HR) pathway, which is active specifically in S- and G₂-phase cells, and by the nonhomologous end-joining (NHEJ) pathway, which can be activated in all cell cycle phases (but has a preference for the G₁ phase). The HR and NHEJ pathways therefore interact with each other across cell cycle transitions to effect DNA-dsb repair and prevent cellular mutation and toxicity in normal tissues (1, 2).

The expression and function of HR and NHEJ proteins may differ between normal and tumor cells, which could provide a therapeutic window for novel therapies (3, 4). Increasingly, multimodality combination treatment (e.g., chemoradiotherapy or radiotherapy with molecular-targeted agents) is being used in many bladder, lung, cervix, and other solid tumors to improve local control and survival. In this regard, correlations have been observed between radiotherapy or chemotherapy outcome and expression or function of DNA-dsb repair proteins. This suggests that an understanding of DNA-dsb repair across the malignant axis could give rise to prediction of therapy response and judicious choice of molecular-targeted agents in combination with chemotherapy and radiotherapy as tumor cell sensitizers (2, 4–6). Novel strategies with HR repair pathways could enhance tumor cell kill based on an increased fraction of cells in S and G₂ phases of the cell cycle, altered DNA damage checkpoints (based on p53 mutation), synthetic lethality with poly(ADP-ribose) polymerase inhibition, and the presence of intratumoral hypoxia, which may modify DNA-dsb repair (2, 7, 8). HR-defective cells are more sensitive to certain agents [e.g., mitomycin C (MMC), cisplatinum, gemcitabine, ionizing radiation (IR), poly(ADP-ribose) polymerase inhibitors] compared with NHEJ-defective cells, suggesting an increased specificity for targeted therapeutics if the underlying defect in DNA-dsb repair is understood (9–11). Indeed, these differences in HR capacity can be assayed by determining the relative sensitivities to cross-linking and alkylating agents such as gemcitabine and MMC, which are also among the most potent radiosensitizers when combined with radiotherapy (12, 13).

A specific HR-related target for cancer therapy is the RAD51 protein. RAD51 protein expression has been linked to MMC, IR, and gemcitabine resistance (14). Recent data also suggest that RAD51 overexpression leads to a worse clinical outcome in lung cancer (15). RAD51 expression can be manipulated pharmacologically in vitro by the use of imatinib mesylate. Imatinib is an inhibitor of c-ABL, c-KIT, and platelet-derived growth factor receptor tyrosine kinases. However, c-ABL also up-regulates Rad51 gene expression and increases RAD51 protein phosphorylation at Tyr315, resulting in increased RAD51 focus formation in a complex with the c-ABL or RAD52 and ATM proteins (16, 17). Treatment of leukemia, prostate, and glioma cells with imatinib can decrease RAD51 expression and sensitize them to experimental chemotherapy and radiotherapy in vitro (16, 18). However, whether imatinib directly decreases error-free HR as a mechanism for this increased tumor cell chemosensitivity or radiosensitivity is not known. If true, this would give unique information about the judicious selection of agents to combine with imatinib and potential biomarkers relating to the use of imatinib in combination with HR-sensitizing agents.

Herein, we have studied the effects of imatinib on tumor cell lines of varying p53 status and histopathology in combination with agents that may preferentially be toxic to HR-defective cells (e.g., IR, MMC, and gemcitabine). We observed increased cell kill in tumor cells in vitro compared with fibroblasts. This was associated with reduced RAD51 expression and function. These findings were recapitulated in a prostate xenograft model treated with radiotherapy and imatinib without attendant gut toxicity. Together, these results support the targeting of HR with imatinib as a cancer therapy that preserves the therapeutic ratio.

The growth characteristics and culture of the cell lines used in this study have been previously described (19, 20). GM05757 (normal diploid fibroblasts), PC3 (prostate adenocarcinoma), and PANC1 (pancreatic adenocarcinoma) were used in these experiments and were obtained from Coriell Cell Repository or American Type Culture Collection. RT112M (transitional bladder cell carcinoma) was a generous gift from Prof. Maggie Knowles. The H1299-DR-GFP human lung carcinoma cell line was a kind gift of Dr. Simon Powell (Washington University, St. Louis, MO) and expresses an integrated direct repeat-green fusion protein (DR-GFP) HR reporter system as previously described (8).

The p53 gene status of each cell strain was determined by direct DNA sequencing of exons 2 to 11 using automated DNA sequencing (DNA Sequenator, Becton-Coulter) as previously described (21). The GM05757 strain and RT112 cell line express two wild-type p53 alleles and functional wild-type p53 protein, whereas the PC3, H1299, and PANC1 cell lines express truncated p53 alleles leading to a lack of expression of the wild-type p53 protein (i.e., null p53). All cells were incubated in 5% CO₂ at 37°C. RT112 cells were grown in RPMI 1640 supplemented with 10% FCS and 1% l-glutamine; H1299 in RPMI 1640, 10% FCS, and 20 mmol/L HEPES; PC3 in F12K with 10% FCS; PANC1 in Dulbecco's and 10% FCS; and NDF-GM05757 in α-MEM with 15% FCS. All media contained broad-spectrum antibiotics (100 mg/L penicillin V and 100 mg/L streptomycin). All cell lines were negative for Mycoplasma contamination.

Whole cell lysates from asynchronously growing and synchronized G₀-G₁ cell cultures were prepared for SDS-PAGE/Western blot analyses (20). Briefly, nonirradiated and irradiated cultures were lysed on ice in lysis buffer [150 mmol/L NaCl, 0.1% SDS, 1% sodium deoxycholate, 1% Triton X-100, 10 mmol/L Tris-HCl (pH 7.4), plus Complete Protease Inhibitor (Roche)] and 20 to 50 μg of protein were equally loaded for SDS-PAGE. Gels were electroblotted onto nitrocellulose membranes (S&S) using semi-dry transfer (Helixx). Membranes were placed in blocking solution (TBST/10% low fat milk) before incubation with the primary antibody of interest. The anti–c-ABL monoclonal antihuman antibody was obtained from Cell Signaling Technology; the anti-Ku70 and anti-RAD51 antibodies from Santa Cruz Biotechnology, Inc.; and the anti-actin antibody from Sigma-Aldrich, Inc. Membranes were then washed with TBST, incubated with the appropriate horseradish peroxidase–conjugated secondary antibody (Santa Cruz Biotechnology and Oncogene Research Products), washed again in TBST, and processed for chemiluminescence detection using the Renaissance chemiluminescence detection kit (NEN). Membranes were then exposed to enhanced chemiluminescence film (Amersham).

For selected experiments, we isolated insoluble nuclear biochemical fractions (chromatin fractions) using a modified Dignam method (22). Briefly, cells were lysed and incubated for 5 min on ice in cytoplasmic buffer [25 mmol/L KCl, 5 mmol/L MgCl₂, 10 mmol/L Tris-HCl (pH 8.0), 0.5% NP40, 1 mmol/L DTT, 1× protease inhibitors (Complete EDTA-free, Roche), 1× phosphatase inhibitors (Cocktail Set II, Calbiochem)]. Lysates were centrifuged at 3,000 rpm for 5 min and the supernatant (cytoplasmic fraction) was separated. The nuclear pellet was rinsed thrice with cytoplasmic buffer, then resuspended in nuclear buffer [10 mmol/L Tris-HCl (pH 8.0), 500 mmol/L NaCl, 1.5 mmol/L MgCl₂, 0.1% NP40, 5 mmol/L EDTA, 1× protease inhibitors (Complete EDTA-free, Roche), 1× phosphatase inhibitors (Cocktail Set II, Calbiochem)]. The nuclei were lysed by vigorous pipetting, incubated on ice for 15 min, and then centrifuged at 3,000 rpm for 5 min. The supernatant was separated and the remaining pellet was rinsed and resuspended in nuclear buffer before sonication. The insoluble nuclear fraction was analyzed by Western blotting. The expression of actin, GRB2, or p84 protein was used as loading controls for total, cytoplasmic, or nuclear fractions, respectively.

Cellular apoptosis was scored as previously described (21) using a nuclear morphology end point. Both adherent and floating cells were collected from cell cultures following treatment with imatinib and gemcitabine or radiation and stained with 10 μmol/L Hoechst 33342 dye in 4% formalin-PBS for 30 min. Stained cells were visualized under a fluorescent microscope to look for distinct apoptotic bodies and nuclear fragmentation.

Cell cycle distributions pretreatment and posttreatment were quantitated by flow cytometry. Treated cells were collected, washed, and fixed in 70% ethanol. The cells were washed with magnesium- and calcium-free PBS before staining with 10 μg/mL propidium iodide and addition of 10 μg/mL DNase-free RNase. Flow cytometry was done using a FACSCalibur flow cytometer (BD Biosciences), and cell cycle distribution profiles were analyzed using CellQuest software (BD Biosciences).

To evaluate error-free HR repair following imatinib treatment, we used H1299 cells containing a DR-GFP construct as previously described (8). In this system, transient expression of an I-SceI endonuclease generates a DNA-dsb at integrated GFP-deletion gene sequences in which error-free repair leads to a full-length GFP product detectable by flow cytometry. H1299 cells were therefore transfected with pGFP (transfection efficiency control) or pCMV3xnlsI-SceI (functional endonuclease) or phCMV-1 I-SceI (negative control) and in the presence or absence of imatinib. GFP signal was quantitated at 3 d posttransfection on a FACSCalibur flow cytometer (BD Biosciences). For each experiment, 50,000 cells were scored per treatment group and the frequency of recombination events was calculated from the number of GFP-positive cells divided by the number of cells analyzed following correction for transfection efficiency as previously described (8).

Rad51-Validated Stealth RNAi DuoPak and the negative oligo (Invitrogen) were used at a concentration of 0.10 nmol/L for a 24-h transfection to obtain ∼50% down-regulation of RAD51 in the cells. Transfection of siRNA or negative oligo was achieved using Lipofectamine 2000 reagent (Invitrogen) as previously described (8).

Cells were irradiated with doses between 0 and 10 Gy under aerobic conditions, at room temperature, using a ¹³⁷Cs unit (Nordion Gammacell) at a dose rate of 1.03 Gy/min, as previously described (23). In other experiments, cells were treated with gemcitabine (Lilly; 0–100 nmol/L), MMC (0–750 nmol/L), or imatinib (Novartis; 0–50 μmol/L). Imatinib was reconstituted in acidified PBS (pH 5.3) to a stock concentration of 10 mmol/L and stored at −20°C. Gemcitabine and MMC were obtained at stock concentrations of 0.127 mol/L and 1.5 μmol/L and stored at room temperature and −4°C, respectively. Asynchronously growing cultures were washed with PBS, trypsinized, and plated at appropriate density into six-well tissue culture dishes before irradiation or treatment with chemotherapy. Dishes were treated for 16 to 18 h after plating to ensure a cell multiplicity of <1.1 (23). At 10 to 14 d posttreatment, dishes were stained with methylene blue-ethanol, and surviving colonies of >50 cells were scored under a dissection microscope. The surviving fraction for a given treatment dose was calculated as the relative plating efficiency of treated versus untreated (control) cultures. The ICD₅₀ was defined by the concentration of drug that results in 50% clonogenic cell survival. For each cell line, three to six independent experiments were done.

Cells were fixed and permeabilized as previously described (24). Cells were stained with mouse monoclonal γ-H2AX antibody (Upstate) at 1:500 dilution or RAD51 antibody (Cell Signaling) at 1:200 dilution and counterstained with Alexa 488–conjugated antimouse secondary at 1:500 dilution (Molecular Bioprobes). DNA staining was accomplished with 4′,6-diamidino-2-phenylindole before microscopy. Cells were examined using either an Axiovert wide-field microscope under control of Northern Eclipse software or a Zeiss two-photon confocal microscope under control of LSM software as previously described (19).

To derive PC3 prostate xenografts for in vivo growth delay studies of combined treatments with imatinib and radiotherapy (IR), male BALB/c nude mice were injected s.c. in the flank with 1.5 × 10⁶ cells. Drug and radiation treatments were started when xenografts reached a volume of about 150 to 200 mm³. Mice were randomly divided into four groups of 9 to 10 mice adjusted by initial tumor size and treated with (a) PBS alone, (b) imatinib 50 mg/kg i.p. daily for 8 d, (c) 5 × 4 Gy plus 8 d PBS i.p., and (d) 5 × 4 Gy plus 8 d of imatinib (IR starting after 3 d of pretreatment with imatinib). Animals were weighed and flank tumors measured twice a week with calipers; measurements were converted into tumor volume. Growth delay was calculated as the time difference (in days) between treatment and control groups to grow to 250 mm³. In a separate experiment, selected tumors were treated similarly and removed 4 h following a single dose of 4 Gy and prepared for Western blotting and paraffin-fixed sections for immunohistochemical staining for RAD51 expression (25).

Gut toxicity was determined using intestinal clonogenic survival in vivo using a modified protocol (26). Briefly, doses of whole-body irradiation at 0 to 18 Gy were administered to BALB/c mice following 50 mg/kg/d × 5 d of imatinib or PBS i.p. dosing. Three days later, the small intestine was removed, washed, and fixed in formalin; gut cross sections were stained with H&E before scoring surviving crypts in each treatment group. Two independent experiments were completed. Crypt radiation surviving fraction in vivo was calculated as the ratio of surviving crypts in the presence and absence of imatinib. All animal experiments were carried out in accordance with the Ontario Cancer Institute/Princess Margaret Hospital Animal Care Committee–approved protocol.

Unless otherwise stated, quantitative experiments were conducted with a minimum of three independent experiments and data expressed as mean ± SE. Significant differences between experimental groups were compared using one-sided or paired t tests.

We initially investigated whether RAD51 protein expression could be decreased by imatinib (Fig. 1; Supplementary Fig. S1).³

A concentration of 10 or 20 μmol/L imatinib (reflecting an inhibition of 50% survival; ICD₅₀) significantly decreased RAD51 protein expression in a time-dependent manner in all cell lines by 40% to 50% (P < 0.05; Fig. 1A). RAD51 expression at 72 hours was markedly lower than at earlier time points independent of p53 status or histologic cell type. In contrast, there was no effect on Ku70 with similar treatment (Fig. 1A). Although RAD51 is differentially expressed in the S and G₂ phases of the cell cycle (reviewed in ref. 4), the imatinib-mediated decrease in RAD51 was not be explained by altered S-phase fraction because cell cycle distributions remained unchanged following imatinib treatment (Supplementary Fig. S1C).³

The coimmunoprecipitation of c-ABL/RAD51 complexes following radiation was also decreased by imatinib treatment (Fig. 1B). Given the effect of imatinib on RAD51 levels, we investigated the effect of imatinib on the pool of chromatin-bound RAD51 by fractionating imatinib-treated cells into a nuclear-insoluble protein pool (Fig. 1C; Supplementary Fig. S1A and B)³ and by observing RAD51 subcellular localization using immunofluorescent microscopy (Fig. 1D). We observed that imatinib significantly decreased RAD51 protein in the chromatin-bound fraction in two of the three irradiated tumor cell lines (P < 0.05) with a strong trend to decrease in the remaining cell line (Fig. 1C; Supplementary Fig. S1A and B).³ Microscopy experiments showed increased nuclear extrusion of RAD51 in imatinib-treated cells (Fig. 1D). We conclude that treatment with ICD₅₀ concentrations of imatinib leads to decreased RAD51 nuclear expression in tumor cell lines in a cell cycle–, p53-, and histology-independent manner.

To determine whether the decrease in nuclear RAD51 protein expression led to altered error-free HR, we used a specialized H1299 lung cancer cell model that contains an integrated DR-GFP HR reporter system (8). The reporter construct contains a GFP gene, which is interrupted by a rare restriction site (I-SceI), and a downstream GFP fragment. Cells were transfected with a plasmid encoding for either a negative control, GFP (transfection efficiency control), or the endonuclease to generate a specific DNA-dsb. In this assay, functional GFP can only be restored if the DNA-dsb is repaired in an error-free manner using the downstream GFP fragment as a template for HR (see Fig. 2A; refs. 8, 27). We tested the HR frequency in H1299 cells following treatment with imatinib (see Fig. 2A and B). The results in Fig. 2B showed that RAD51 protein was significantly reduced to 40% to 70% following treatment with 10 or 20 μmol/L of imatinib (P < 0.05). In a parallel assay, 10 and 20 μmol/L of imatinib decreased the frequency of recombination to 62% and 77%, respectively (P < 0.05), supporting a concordance between RAD51 expression and HR following imatinib treatment.

We compared this effect of imatinib on RAD51 expression and HR with the effect of RAD51 knockdown in the same cell line using siRNA treatment. RAD51 knockdown led to a 40% to 50% decrease in RAD51 protein expression (similar to that achieved by imatinib treatment; compare Fig. 2B and C), reduced HR by 30% to 40%, and increased sensitivity to IR and MMC. We conclude that imatinib and RAD51-siRNA are both capable of inhibiting HR in the H1299/DR-GFP cellular model.

Given the decreased RAD51 expression and function in selected tumor cell lines, we determined whether the ability of imatinib to render tumor cells deficient in RAD51 expression led to increased tumor cell increased sensitivity to IR, MMC, or gemcitabine, agents that may selectively be toxic to DNA-dsb repair–deficient cells (2, 4, 7, 23). We observed that imatinib also reduced RAD51 expression following an initial IR or gemcitabine treatment in vitro (see Fig. 3A and B). Tumor cells treated with imatinib and IR, MMC, or gemcitabine had significantly decreased clonogenic survival (P < 0.05) compared to cells treated with IR or chemotherapy alone (Fig. 3B and C). The effect of imatinib as a chemosensitizer was more pronounced in tumor cells compared with normal diploid fibroblasts (P < 0.05; Fig. 3D).

Given the results of the clonogenic assays, we characterized the expression of the DNA-dsb biomarker γ-H2AX in imatinib-treated cells. Using an apoptotic assay, we found no evidence that imatinib altered the apoptotic response in the cell lines treated with IR or chemotherapy. For example, the AI for RT112 cells with radiation was 8% after 10 Gy at 72 hours and 6% after treatment with gemcitabine at 72 hours (data not shown). However, treatment with the combination of gemcitabine or IR and imatinib resulted in increased residual γ-H2AX signal and mitotic catastrophe (see Fig. 4). We conclude that imatinib can sensitize tumor cells to DNA-dsb–inducing agents and induce mitotic catastrophe.

To support the in vitro imatinib-IR sensitization data in PC-3 cells, we used a PC3 prostate xenograft model to test whether we could observe radiosensitization in vivo following a combination of imatinib and experimental fractionated radiotherapy (Fig. 5). We also tested whether the combination of radiation plus imatinib would be toxic in an associated normal tissue pelvic irradiation model based on intestinal crypt cell survival.

Figure 5A shows that imatinib at 50 mg/kg/d for 5 days did not lead to increased radiation gut toxicity. Studies using the PC-3 xenografts showed decreased RAD51 protein expression in vivo based on Western blotting or immunohistochemical staining of RAD51 protein of PC3 cells within irradiated xenografts (Fig. 5B and C). The combination of imatinib and fractionated radiotherapy (five fractions of 4 Gy) led to an increased growth delay (100 ± 7 days) compared with radiotherapy alone (74 ± 14 days; P = 0.003; Fig. 5D). There was no significant growth delay in the imatinib-alone treatment group. Over the duration of the growth delay experiment, we did not observe significant differences in body weight among the treatment groups (Fig. 5D).

Imatinib has been used in the clinic for a number of years in the setting of chronic myelogenous leukemia and rare gastric sarcomas and is considered to be safe and well tolerated (24). Previous studies have shown that imatinib can reduce the levels of cellular RAD51, a pivotal protein involved in DNA-dsb repair, leading to enhanced radiosensitization (28, 29). However, whether this was secondary to functional decreases in HR was not studied. The latter is important to choose the most effective therapies to combine with imatinib and lead to the development of additional biomarkers relating to HR. Our data suggest that imatinib can decrease functional HR leading to increased chemosensitivity and radiosensitivity in vitro and in vivo without undue toxicity toward normal cells and tissues; we believe that this effect is, in part, optimizing the therapeutic ratio for these combined treatments.

Imatinib can inhibit c-ABL, platelet-derived growth factor receptor, mitogen-activated protein kinase, and c-KIT signaling pathways to decrease interstitial fluid pressure and angiogenesis in vivo, but there are also reports of altered DNA repair (30–38). Additionally, c-ABL has also been shown to be involved in the transcription and posttranslational modification of RAD51 (17). Because the platelet-derived growth factor-mitogen-activated protein kinase pathway is not always intact in malignant cell lines, our results suggest that our observed increases in cytotoxicity in vitro and in vivo following combined treatment with imatinib and DNA damaging agents may be due, in part, to inhibition of DNA repair (30, 31). Indeed, initial Western blotting for phosphorylated platelet-derived growth factor receptor and mitogen-activated protein kinase among the tumor cell lines in this study following treatment with imatinib was variable from cell line to cell line (data not shown). In contrast, RAD51 was uniformly reduced in all cell lines after 72 hours. This was not due to differential mRNA expression of the Rad51 gene (data not shown) and may therefore be secondary to imatinib effects on RAD51 translation or posttranslational modification. A further mechanism for reduced HR could stem from our observation that RAD51 expression was specifically inhibited within the cell nucleus rather than the cytoplasm. Recent reports confirm that intracellular trafficking of the RAD51 protein and functional HR can be modified by insulin-like growth factor type I receptor or hepatocyte growth factor (MET) signaling pathways (39, 40). Future research using defined RAD51 or signaling isogenic systems will be required to test whether these mechanisms are operational in imatinib-treated cells.

In the presence of decreased RAD51, IR or chemotherapy combined with imatinib was shown to significantly decrease tumor cell survival. Ku70 levels were not affected by imatinib in this study. This suggests that a differential effect between HR and NHEJ could be exploited in radiotherapy protocols because these pathways may be differentially used in malignant cells and normal cells, respectively (reviewed in ref. 4). Indeed, the survival of the normal fibroblasts was higher following imatinib treatment in comparison with tumor cells, and we speculate that this is due to the increased number of cells in the G₁ phase in the GM05757 strain (∼85%), which is preferential to NHEJ-directed DNA repair. The increased cell kill in tumor cells would be secondary to higher S and G₂ fractions in which HR inhibition would be maximal (4).

Previous studies have shown enhanced radiosensitization in glioblastoma cell lines and xenografts (29, 34, 35). Podtcheko et al. (36) have shown that imatinib enhanced radiosensitivity in anaplastic thyroid cells both in vitro and in vivo. They concluded that not only inhibition of proliferative cell growth but also terminal growth arrest and senescence were important (36). Senescence refers to irreversible arrest and loss of clonogenicity. Mitotic cell death occurs in cells with abnormal cell cycling and defective apoptosis with irradiation or chemotherapy (37). A recent study by Rink et al. (32) has implicated NBS1, a key protein in the MRE11-NBS1-RAD50 complex involved in DNA-dsb signaling and cell cycle control, in leukemic cell killing following imatinib treatment (33). In this study, the treatment resistance of BCR/ABL-positive leukemic cells was reversed by imatinib through inhibition of NBS1 and cell cycle arrest, leading to decreased capacity for DNA repair following DNA damage. Whether our combined treatment results can be explained by effects on the NBS1 pathway will require specific study, but they are consistent with our observation of increased mitotic catastrophe following imatinib treatment in combination with chemotherapy or radiotherapy.

Recently, Bertino et al. (38) investigated the combination of gemcitabine or premetrexed with imatinib in mesothelioma cells and found the combinations to be synergistic. Gemcitabine is used in the treatment of lung, pancreas, and bladder cancer. Interestingly, in this study, these cell lines showed a greater effect with combination gemcitabine and imatinib treatment when compared with the prostatic cell line. Although further study is required with multiple cell lines within each histopathologic category in vitro and in vivo, we cautiously suggest that the combination may be optimal in patients with lung, pancreas, or bladder cancer. In the United Kingdom, a phase II trial of concurrent gemcitabine with radiotherapy in the treatment of bladder cancer (radiation therapy weekly with gemcitabine)⁴

has just closed and results are awaited with interest. If the gemcitabine and radiotherapy combination is shown to be more efficacious than radiotherapy alone in a phase III trial, then future studies may look at the outcome of trimodality treatment using radiation, chemotherapy, and imatinib, with possible further improvements in clinical outcome and reduced toxicity compared with current treatments. Finally, given the increasing data that RAD51 overexpression is associated with malignancy and can lead to chemoresistance and radioresistance (41), HR-related biomarkers such as RAD51 nuclear expression could potentially be used as biomarkers of efficacy in trials testing HR-targeted therapies (42).

The authors received grant support from AstraZeneca-KuDOS, Xceed, and SuperGen for preclinical research. No other potential conflicts of interest were disclosed.

We thank Carla Coackley for her technical expertise.