Imatinib Radiosensitizes Bladder Cancer by Targeting Homologous Recombination

Bladder cancer is the fourth most common cancer in men in the United Kingdom (1). In a population-based study, radiotherapy was found to be as effective as cystectomy in the treatment of muscle-invasive disease and is being increasingly required as the population ages (2). Radiotherapy to the bladder results in acute bladder and bowel toxicities in most patients and, more rarely, causes long-term toxicity in which the most severe cases may require a cystectomy for alleviation of symptoms. Conventional cytotoxic chemotherapy agents have been used to improve the outcome of radiation treatment in muscle-invasive bladder cancer (3–5). However, elderly patients are not always able to tolerate conventional chemotherapy agents when used as radiosensitizers.

For tumors with genomic aberrations/alterations, therapies targeted toward the expressed proteins, such as the tyrosine kinase inhibitors (TKI) imatinib and lapatinib, are an attractive option as they do not have the myelosuppressive or neurotoxic side effects of chemotherapy, although they do cause diarrhea, skin rash, and very rarely lung fibrosis, and most are available as oral preparations. However, it is important that such costly agents are targeted to those patients most likely to benefit. Imatinib selectively inhibits the tyrosine kinase activity of ABL as well as several receptor tyrosine kinases: the platelet-derived growth factor receptors (PDGFR-α and -β), the stem cell factor (SCF) receptor (KIT), the discoidin domain receptors (DDR1 and DDR2), and the colony-stimulating factor receptor (CSF-1R; refs. 6, 7) and is used to treat chronic myelogenous leukemia (CML) and gastrointestinal stromal tumors (GIST). There are currently no clinical trials involving imatinib in bladder cancer. Lapatinib is a TKI that selectively targets the EGF receptor (EGFR) and HER2 (8) and is indicated in the treatment of HER2-positive breast cancer. Clinically, lapatinib is primarily being investigated in metastatic bladder cancer. However, there is an ongoing pilot study of neoadjuvant lapatinib before cystectomy (9).

Bladder tumors express tyrosine kinases to varying extents: 50% and 45% overexpress EGFR and HER2, respectively (10), whereas PDGFR was reportedly expressed in approximately 80% in a Chinese cohort, published only in Chinese. Although KIT expression has been described in both upper tract tumors and small cell carcinomas of the bladder (11, 12), neither KIT nor ABL expression has been studied in transitional cell carcinomas of the bladder to our knowledge.

Ionizing radiation causes DNA damage, including base damage, single-strand breaks (SSB), and double-strand breaks (DSB). Unrepaired or misrepaired DSBs are lethal, resulting in cell death both in tumors and normal tissues (13, 14). Mammalian cells use 2 major pathways to repair DSBs, namely homologous recombination and nonhomologous end-joining (NHEJ; refs. 15, 16). Homologous recombination is an error-free pathway, predominantly used in the G₂–S phase of the cell cycle, which requires the sister chromatid to act as a DNA template. A major homologous recombination protein is RAD51, which is involved in ATP-dependent DNA strand exchange. RAD51 expression is increased following ionizing radiation in tumor cells and induces the formation of RAD51 nuclear foci at sites of DSB (17). NHEJ is the major DSB repair pathway used in G₀ and G₁ (18) and involves the DNA-binding complex Ku70/Ku80 and the DNA-dependent protein kinase catalytic subunit (DNA PKcs; ref. 19). In addition to the classical DSB repair pathways, a less efficient Ku-independent pathway has been described in Ku-deficient yeast cells, which involves microhomology-mediated end-joining (MMEJ; ref. 20).

We have previously shown the MMEJ phenotype in vitro using cell-free extracts from muscle-invasive bladder tumors, and this is associated with reduced Ku–DNA binding and loss of TP53 function (21). Pucci and colleagues also showed reduced Ku–DNA binding in 5 advanced breast and muscle-invasive bladder tumors (22). This error-prone repair was not detected in normal human urothelial cell extracts (23), which suggests a therapeutic window that could be targeted by novel therapies. Negroni and colleagues (24) inhibited Ku80 expression by RNA interference (RNAi) using short hairpin RNA (shRNA) in RT112 bladder cancer cells and showed increased radiosensitivity and reduced Ku–DNA binding compared with parental RT112 vector-transfected cells.

The outcome of clinical radiotherapy depends, in part, upon the extent of DNA damage and how efficiently the cells can repair this damage. Selective targeting of DNA repair pathways could increase tumor cell kill while sparing normal tissues, thus increasing the therapeutic ratio. ABL upregulates RAD51 gene expression and imatinib reduces RAD51 protein expression and RAD51-chromatin binding (25) and reduces error-free homologous recombination efficiency (17). Imatinib also reduces the increased RAD51 expression induced by ionizing radiation and reduces the associated RAD51 nuclear focus formation in glioma cell lines (25), and in bladder, pancreatic, prostate, and lung carcinoma cell lines, imatinib increases cell kill in combination with ionizing radiation, due in part to mitotic catastrophe (17), unlike normal fibroblasts, where cell survival is unaffected. In xenograft studies, imatinib increases growth delay following fractionated radiotherapy in glioblastoma, epidermoid, and prostate carcinoma models with no apparent increase in toxicity (17, 26).

EGFR inhibitors, such as lapatinib, can act as radiosensitizers by targeting the intracellular signaling cascades (including Ras-MAPK and PI3K-AKT), which are triggered by binding of ligand to the transmembrane EGFR ligand-binding domain (27, 28). These cascades are normally activated by ionizing radiation in EGFR-overexpressing tumors, resulting in radioresistance (27). Moreover, they also act as radiosensitizers by repressing DNA repair in irradiated cells although this occurs via the NHEJ pathway, through inhibition of the phosphoinositide 3-kinase (PI3K)–mediated stimulation of DNA PKcs and by blocking of the nuclear interaction between EGFR and DNA PKcs normally induced by ionizing radiation [reviewed by Baumann and colleagues (27)].

We hypothesized that in muscle-invasive bladder cancer, it would be better to use an agent that targets the homologous recombination pathway rather than the NHEJ pathway. This would result in a form of “synthetic sickness” (see ref. 29 for recent review), where tumor cells already deficient in NHEJ would have reduced homologous recombination efficiency and repair, and thus increased ionizing radiation–induced lethality. As imatinib is known to target homologous recombination via RAD51 (17, 25), it may be one such agent, and the combination of imatinib and radiotherapy should result in an increased therapeutic ratio for muscle-invasive bladder cancer.

We therefore sought to see whether ABL, KIT, PDGFR, HER2, or EGFR are targets in bladder cancer by determining their role in radiotherapy response. We then moved to an experimental system to test the effectiveness of therapies that might be used in combination with radiotherapy to enhance radiosensitivity. This included targeting the ABL/RAD51 and EGFR pathways.

Ninety-one formalin-fixed paraffin-embedded bladder tumor biopsy samples were obtained from patients treated with radical radiotherapy for transitional cell carcinoma of the bladder at the Leeds Cancer Centre (West Yorkshire, United Kingdom) from 2002 to 2005. Details of the patients and radiotherapy treatments have been described previously (30). Patients gave informed consent for use of their tissues and local ethical approval was obtained from the Leeds (East) Research Ethics Committee (project 04/Q1206/62).

Antibody conditions were optimized by staining sections from control tissues, namely breast tumor for ABL, HER2, EGFR, and PDGFR and skin tumor for KIT. Then 4-μm sections from the bladder tumor specimens were heated, dewaxed, and hydrated in xylene, graded alcohols, and water. Antigen retrieval was achieved by boiling in EDTA (1 mmol/L at pH 8.0) or citric acid (pH 6.0) for 2 minutes of pressure cooking before quenching endogenous peroxidase activity with 3% H₂O₂ for 30 minutes. Endogenous protein-binding activity was blocked using avidin–biotin–blocking agent (Vector) and normal goat serum (Dako) before incubation with primary antibody: anti-ABL (1:250; NeoMarkers), anti-c-HER2 (1:200, Dako), anti-EGFR (1:100; Novocastra), anti-KIT (1:40; Novocastra), and anti-PDGFRβ (1:25; Cell Signalling) diluted in diluent (Dako) for 1 hour. Samples were then incubated for 30 minutes with secondary antibody conjugated to horseradish peroxidase using the Dako ChemMate Detection Kit (Dako). Immunoreactivity was revealed by incubation of sections with 3-3′-diaminobenzidine (DAB) for 10 minutes before washing, taking through graded alcohols, clearing in xylene and counterstaining with hematoxylin (VWR) before mounting in dibutylphthalate xylene (Leica). Digital images were captured within invasive tumor areas (3–10 images per slide, ×400 magnification) using an Olympus BX50 microscope and c-3030 camera.

Assessment was made of the percentage and intensity of tumor cells with membranous staining based on the recommendations for interpretation of the HercepTest: gastric cancer (31). Briefly, for all antibodies studied, complete, basolateral, or lateral membrane staining was scored as an intensity of 3+ for strong intensity and 2+ for weak to moderate intensity, where at least 10% of the tumors cells stained positive. Partial membrane staining or only faint/barely perceptible intensity staining in at least 10% of tumor cells was scored as a 1+. Staining intensity was scored independently in a blinded manner by 2 observers, discordant scores were reviewed, and a consensus was reached.

Cause-specific survival was defined from day 1 of radiotherapy until death from bladder cancer. Death from another cause was considered a censored observation. Kaplan–Meier curves were plotted for cause-specific survival and the log-rank statistic used to compare survival times across categories of protein expression.

Imatinib was a generous gift from Novartis Pharma AG (Switzerland) and was later purchased from Stratech Scientific Ltd.; lapatinib was a generous gift from GlaxoSmithKline plc. For cell culture experiments, imatinib and lapatinib were dissolved in dimethyl sulfoxide (DMSO; Sigma) to a stock concentration of 10 mmol/L and stored in single-use aliquots at −20°C.

The TP53 wild-type RT112 bladder transitional cell carcinoma cell line has been authenticated in M.A. Knowles's laboratory by extensive genomic analysis [microsatellite typing, conventional karyotypic analysis, multiplex FISH (MFISH), array-based copy number analysis]. Cells were grown in RPMI-1640 (Sigma) supplemented with 10% v/v FBS (Sigma) and 2 mmol/L l-glutamine (Sigma) in a humidified atmosphere containing 5% CO₂ at 37°C. Exponentially growing cells were used in all experiments.

Cells were harvested from exponential-phase cultures and diluted to 1,000 cells/mL. Five milliliter cell suspensions were then irradiated in tubes at a dose-rate of 1.0 Gy/min using an X-ray machine (Irradiator 320, NDT Equipment Services Ltd.), or cesium-137 source at 1.12 Gy/min using a Gamma-Service Medical GmbH GSR D1 irradiator. The cells were then replated into 10-cm dishes at appropriate cell densities.

Exponentially growing cells were incubated in 75 cm² flasks for 24 hours at appropriate drug concentrations (specified in the figure legends), and the cells then trypsinized and resuspended in medium containing drug as required.

Following the necessary treatments, cells were plated at appropriate cell numbers in triplicate in 10-cm culture dishes containing 10 mL of fresh medium, with or without drug as required (see relevant figure legends). After 14 days incubation, the cells were stained with 1% methylene blue (Sigma) in 50% ethanol, and colonies with more than 50 cells were counted. The surviving fraction was determined as the total number of colonies formed divided by the total number of cells plated multiplied by the plating efficiency, as determined in untreated cells. Radiation survival curves were plotted after normalization for the cytotoxicity induced by control or drug alone, in GraphPad Prism, using the linear-quadratic model with the equation SF = exp −(αD + GβD²). Each point on the survival curve represents the mean surviving fraction from at least 3 independent experiments.

The siRNA expression vector pSilencer 2.1-U6 neomycin (Ambion), which contains a human U6 RNA polymerase III promoter able to transcribe shRNAs, was used in these experiments. Two constructs were made for each of Ku80 and RAD51 (see Supplementary Table S1 and Supplementary Fig. S1), whereby complementary oligonucleotides were used to encode hairpin siRNA inserts, designed to target a 21-mer sequence of human Ku80 coding region or 3′-untranslated region (UTR) mRNA, or RAD51 coding region mRNA, respectively. These were designed using the Ambion Insert Design Tool for pSilencer vectors and purchased from Sigma, and were then annealed and ligated into the linearized pSilencer vector. Circular negative control pSilencer neovector, that expresses a hairpin siRNA with limited homology to any known sequence in the human genome, was used in experiments as a nonspecific negative control.

The target sequences in the human Ku80 or RAD51 gene were determined empirically and analyzed by BLAST search, to confirm a lack of homology to other coding sequences, as per manufacturer's recommendations. The target sequences of clones used for clonogenic assays were Ku80 5′-AAC TCC ATT CCT GGT ATA GAA-3′ (Ku80 coding region target sequence 1) and for RAD51 5′-AAT CAC TAA TCA GGT GGT AGC-3′ (RAD51 coding region target sequence 2; Supplementary Table S1). The corresponding targeting oligonucleotide sequences were for Ku80: top strand 5′-GAT CCG CTC CAT TCC TGG TAT AGA ATT CAA GAG ATT CTA TAC CAG GAA TGG AGT TTT TTG GAA A-3′ and bottom strand 5′-AGC TTT TCC AAA AAA CTC CAT TCC TGG TAT AGA ATC TCT TGA ATT CTA TAC CAG GAA TGG AGC G-3′, and for RAD51: top strand 5′-GAT CCG TCA CTA ATC AGG TGG TAG CTT CAA GAG AGC TAC CAC CTG ATT AGT GAT TTT TTG GAA A-3′ and bottom strand: 5′-AGC TTT TCC AAA AAA TCA CTA ATC AGG TGG TAG CTC TCT TGA AGC TAC CAC CTG ATT AGT GAC G-3′.

RT112 cells were seeded into 75-cm² flasks and the following day transfected at 60% confluence with pSilencer-Ku80, pSilencer-RAD51, or circular negative control pSilencer neovector, using Lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. After 24 hours, fresh medium was added containing the selection reagent G418 (600 μg/mL; Gibco, Invitrogen). Selection was continued for 14 days, with the medium refreshed every other day. Single clones were picked and tested for Ku80 and RAD51 expression, respectively, by Western blotting and subsequently tested for ionizing radiation sensitivity using clonogenic assays.

Cells were lysed on ice in radioimmunoprecipitation assay (RIPA) buffer (Sigma) with 1% of protease inhibitor and phosphatase inhibitor (Sigma). The cells were allowed to swell on ice for 20 minutes; then the lysate was centrifuged for 30 seconds at 12,000 × g. The supernatant was carefully removed and stored at −20°C. Total protein concentration in cell lysates was determined by the method of Bradford (Sigma). Thirty to 50 μg of protein was resolved on 4% to 20% polyacrylamide gels and transferred onto nitrocellulose membranes. The resulting membranes were incubated with blocking buffer (Li-cor) and primary antibodies. The antibodies used were rabbit polyclonal ABL (2862; Cell Signalling), rabbit polyclonal Rad51 (ab63801; Abcam), rabbit polyclonal EGFR (sc-03; Santa Cruz Biotechnology), mouse monoclonal anti-Ku80 (Ab-2; Neomarkers), mouse monoclonal β-tubulin (clone TUB2.1 TT4026; Sigma), and mouse monoclonal β-actin (ab8226; Abcam).

Fluorochrome-conjugated secondary antibodies (Li-cor) were used and detected by infra red scanning densitometry using the Li-cor Odyssey Infrared Detection System (Li-cor Biosciences UK Ltd.). Quantification was based on normalization to β-actin. The immunoblotting experiments were carried out at least 3 times.

Propidium iodide (PI) staining and flow cytometry were used to determine cell-cycle stages. Cells from the batches used for clonogenic assays were washed with PBS, fixed in ice-cold 70% ethanol, and stored at 4°C before analysis. Cells were spun down and resuspended in PI solution (50 μg/mL PI, 0.5 mg/mL RNase; Applied Biosystem) and incubated at room temperature in the dark at least for 30 minutes. DNA content was detected by flow cytometry (Beckman FACScan system). The relative proportions of cells in the G₁, S, and G₂–M phases of the cell cycle were determined using ModFit LT 3.2 software.

All statistical analyses were conducted using SPSS16.0 software. Clonogenic assays were conducted in triplicate at least 3 times, with the results expressed as mean + SD as appropriate. Western blot analyses were conducted at least 3 times independently. Results were considered statistically significant at a P < 0.05, using a two-tailed unpaired Student t test. Sensitizer enhancement ratios (SER) were calculated at a surviving fraction of 0.1 (10% survival). Cause-specific survival was defined from day 1 of radiotherapy until death of the patient from bladder cancer. Death from another cause was considered a censored observation. Kaplan–Meier curves were plotted for cause-specific survival and the log-rank statistic used to compare survival times across categories of protein expression.

We sought to estimate the percentage of muscle-invasive bladder tumors expressing the tyrosine kinases of interest (imatinib: ABL, KIT, and PDGFR-β; lapatinib: HER2 and EGFR), on the basis that this is likely to represent the cohort of patients for which the addition of a TKI to their radiotherapy may be beneficial. We also wished to determine whether these factors were prognostic in these patients, who had not received such treatments in addition to radiotherapy. The Dako scoring system for the HercepTest was applied to our tyrosine kinases of interest in 91 formalin-fixed paraffin-embedded bladder tumor samples (Fig. 1A). A sample was classified as positive if there was membranous staining in at least 10% of cells, which met the threshold criteria for intensity of staining (2+ or 3+). In 75 of 91 (82%) cases there was positive immunostaining for ABL. In contrast, KIT expression was undetectable in most of the cases but showed weak staining in 8 cases, and PDGFR-β was expressed in endothelial cells and smooth muscle cells but was undetectable in bladder tumor cell membranes. There were 87 of 91 cases (96%) positive for EGFR staining and 86 of 90 (96%) cases HER2-positive.

We also correlated tyrosine kinase expression with patient survival, to look for prognostic significance of high expression. We classified patients into those with low tumor tyrosine kinase expression (equal or less than 2+) and patients with high (3+) expression levels (as there were insufficient tumors scoring 0/1+ for meaningful comparison with 2+/3+). Neither ABL, EGFR, nor HER2 were significantly correlated with patient survival (Fig. 1B).

Before combining drug and radiation treatments, it was first necessary to determine the growth-inhibitory activity of the drugs in RT112 cells. Figure 2 shows cell viabilities by clonogenic assay following 14 days treatment with varying concentrations of imatinib or lapatinib. The IC₅₀ value for imatinib in RT112 cells was 8 μmol/L and for lapatinib 3.5 μmol/L.

To generate NHEJ-deficient and homologous recombination–deficient RT112 cells, we used a siRNA-based strategy to reduce Ku80 or RAD51 expression in RT112 cells. We tested the effectiveness of 2 21-mer siRNAs targeting different sites within the exons or 3′-UTR regions of Ku80 and RAD51 (Supplementary Table S1), using pSilencer2.1-U6 neomycin, which drives expression of a shRNA from the human U6 promoter. The hairpin RNA is then processed into an siRNA, which induces RNAi of the target gene. RT112 cells were stably transfected with each of the vectors and clones picked and tested using Western blotting and ionizing radiation clonogenic assays. An individual clone for each of Ku80 (C13, coding region) and RAD51 (795J) was then selected for drug-ionizing radiation clonogenic assays. Densitometric analysis of representative Western blot analyses revealed a 30% reduction in Ku80 protein expression in C13 cells and 75% reduction in RAD51 expression in 795J cells compared with parental RT112 cells and pSilencer neovector negative control cells (Supplementary Fig. S1).

We examined the radiosensitizing effect of siRNA-mediated downregulation of Ku80 and RAD51 expression in RT112 cells using clonogenic assays. We found a statistically significant increase in radiosensitivity for the Ku80KD (P = 0.01) and RAD51KD (P = 0.02) cells as compared with parental RT112 and RT112-pSilencer vector control cells (Fig. 3).

We then tested our hypothesis that imatinib radiosensitizes NHEJ-deficient Ku80KD cells, whereas having less effect on the parental RT112 cells with an intact NHEJ pathway, due to targeting the ionizing radiation-induced increase in RAD51. We conducted clonogenic assays on RT112, Ku80KD, and RAD51KD cells using relevant drug–ionizing radiation combinations. Cells were incubated for 24 hours with or without drug, before irradiation and plating followed by 14 days incubation, before staining and counting. Drug doses were chosen at approximately IC₄₀ (high) and at IC₁₀ nontoxic (low) levels.

Radiation survival curves were generated for each cell line after normalization for the level of cell killing induced by drug alone. As shown in Fig. 4A–C, 3 μmol/L imatinib had a significant radiosensitizing effect on Ku80KD cells (SER 1.27; P = 0.03) but no significant effect in parental RT112 cells (SER = 1.07; P = 0.051) or RAD51KD cells. At 6 μmol/L concentration, equivalent to approximately IC₄₀, imatinib produced an SER of 1.69 in Ku80KD cells (P = 0.03) but only an SER of 1.34 in the RT112 parental cell line (P = 0.03; P = 0.046 compared with Ku80KD 6 μmol/L imatinib survival curve) and 1.11 in the RAD51KD cells (P = 0.03; P = 0.04 compared with Ku80KD 6 μmol/L imatinib survival curve).

In contrast, lapatinib radiosensitized RT112 cells in a dose-dependent manner (Fig. 4D; SER = 1.13 and 1.30 for 0.6 μmol/L; P = 0.02 and 2.4 μmol/L; P = 0.01) but had no radiosensitizing effect on Ku80KD cells at either low (0.6 μmol/L) or high (2.4 μmol/L) concentrations (Fig. 4E). However, in contrast to imatinib, lapatinib had a dramatic radiosensitizing effect in RAD51KD cells, even at low dose (SER = 1.27; P = 0.02), with the high dose resulting in inhibition of RAD51KD cell colony formation, even without ionizing radiation (Fig. 4F).

To determine whether the drug-mediated enhancement of radiosensitivity was due to cellular synchronization into a radiosensitive phase of the cell cycle, PI staining and flow cytometry were used to determine the cell-cycle phase distribution of samples used in each independent clonogenic assay. Neither imatinib nor lapatinib caused obvious cell-cycle arrest in treated cells of each cell type, except for a small increase in G₁ fraction in RAD51KD cells treated by lapatinib at both drug concentrations (Fig. 5; P = 0.02 low dose; P = 0.04 high dose).

We also measured expression of ABL, EGFR, HER2, and RAD51 using Western blotting (Fig. 6). Cells were incubated with or without drug for 24 hours and then irradiated to 8 Gy or left untreated. Forty-eight hours later, cells were lysed for Western blotting. RT112, Ku80KD (C13) and RAD51KD (759J) cells all had detectable baseline levels of ABL, EGFR, and RAD51 although RAD51 levels were markedly reduced in the RAD51KD cells. HER2 was not detectable in any of the cell lines. We found no major increase in RAD51 levels following 8 Gy ionizing radiation. Imatinib treatment was associated with reduced RAD51 levels in RT112 and Ku80KD cells both alone and following 8 Gy ionizing radiation but no effect on ABL levels; lapatinib had no effect on RAD51 levels.

Radiotherapy is a valid option in the radical treatment of muscle-invasive bladder cancer, with similar survival rates to cystectomy in our recent study (2). Cytotoxic chemotherapy is often combined with radiotherapy to improve survival rates (3–5), but this can be at the expense of late side effects. Moreover, many patients with bladder cancer who elect to have radiotherapy are elderly with poor renal function, and therefore cannot tolerate cisplatin-based chemotherapy regimens. There is therefore an urgent need to identify less toxic agents suitable for these patients.

Our immunohistochemistry data show that ABL is expressed in more than 80% of muscle-invasive bladder tumors, suggesting that imatinib might be useful in such patients. EGFR and HER2 staining was positive in more than 95% of bladder tumors. While the Kaplan–Meier survival curves (Fig. 1B) showed no prognostic significance, our sample numbers were relatively small, with only 73% power to detect a HR of 0.4, with 35 cause-specific survival events at P = 0.05. Also, we used a scoring system developed for another tumor type and have not validated our findings in an independent patient cohort, so results should be treated with caution. As we did not have access to a patient cohort treated with a TKI and radiotherapy, we cannot comment on the potential predictive value of such markers.

We established that both imatinib and lapatinib are cytotoxic in the RT112 bladder cancer cell-line, which has modest ABL and EGFR expression levels (Fig. 2). Choudhury and colleagues (17) found an IC₅₀ value of 20 μmol/L for imatinib in RT112 cells, whereas ours was 8 μmol/L; McHugh and colleagues found the IC₅₀ for lapatinib in RT112 cells was 1.1 μmol/L MTT assay (32), ours being 3.5 μmol/L on clonogenic assay, perhaps reflecting differences in experimental systems. We then successfully knocked-down both Ku80 and RAD51, using shRNA technology, in TP53 wild-type RT112 cell lines (Fig. 3). Negroni and colleagues (24) used the same method to knockdown Ku80 in RT112 cells and showed radiosensitivity and reduced Ku–DNA binding. We previously observed the latter in muscle-invasive bladder tumor extracts.

We then showed the radiosensitizing effects of both imatinib and lapatinib in RT112, Ku80KD, and RAD51KD cells (Fig. 4). Imatinib was an effective radiosensitizer in Ku80KD cells but less effective in parental RT112 cells. Although it did not affect ABL expression levels (Fig. 6) in both parental RT112 cells and Ku80KD cells, imatinib treatment was associated with reduced RAD51 expression levels both with and without ionizing radiation. As imatinib had no radiosensitizing effect at low dose and only limited effects at high dose in RAD51KD cells, this seems to support our hypothesis that imatinib works through the homologous recombination pathway but not the NHEJ pathway. In contrast, lapatinib had no radiosensitizing effect on Ku80KD cells but a marked effect in RAD51KD cells and also sensitized parental RT112 cells, consistent with lapatinib acting via NHEJ rather than homologous recombination. Recently, Myllynen and colleagues (33) showed that both homologous recombination and NHEJ are involved in regulation of DSB repair by EGFR, using an I-SceI–based reporter system, with reduction of homologous recombination by the TKI erlotinib at 0.5 μmol/L and less so by the monoclonal antibody cetuximab at 30 nmol/L concentration. However, our experiments suggest that lapatinib does not act on homologous recombination at 0.6 or 2.4 μmol/L. Other than a small effect on G₁ arrest for lapatinib at both concentrations in RAD51KD cells, neither drug affected cell-cycle progression at 24 hours. Lapatinib has been found to induce G₁ arrest in bladder cancer (RT112, 45%–65% G₁ cells after 72 hours of 1.1 μmol/L lapatinib; ref. 34) and gastric cancer cell lines (58% G₁ to 70% after 24 hours of 1 μmol/L lapatinib; ref. 35). Treatment with imatinib (5–6 μmol/L) for 48 to 72 hours caused a slight increase in the number of cells in G₁-phase in head-and-neck squamous carcinoma cell lines (47% G₁ to 58%; ref. 36), ovarian cancer cell lines (80% G₁ to 94%; ref. 37) but not in small-cell lung carcinoma cell lines (38).

Our data support the role of DNA repair in both imatinib and lapatinib's radiosensitizing effects. They suggest that imatinib may be a useful radiosensitizer in muscle-invasive bladder cancer, as our tumors so far have all shown the MMEJ phenotype, with defective Ku–DNA binding and defective TP53 function, whereas superficial tumors had intact NHEJ (21). We are currently testing further tumor sample extracts. If some muscle-invasive tumors show intact NHEJ, imatinib would not be appropriate as a radiosensitizer in these patients, and there would be a need to develop a preradiotherapy predictive end-joining assay for use in patients with muscle-invasive bladder cancer. An advantage of imatinib as a radiosensitizer in this context, unlike lapatinib, is that we would expect normal tissue sparing and an increase in therapeutic ratio.

No potential conflicts of interest were disclosed.

Conception and design: B. Qiao, M. Kerr, R.M. Phillips, A.E. Kiltie

Development of methodology: B. Qiao, M. Kerr, R.G. Bristow, A.E. Kiltie

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Qiao, A.E. Kiltie

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Qiao, M. Kerr, B. Groselj, M.T.W. Teo, R.M. Phillips, A.E. Kiltie

Writing, review, and/or revision of the manuscript: B. Qiao, M. Kerr, B. Groselj, M.T.W. Teo, M.A. Knowles, R.G. Bristow, R.M. Phillips, A.E. Kiltie

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Qiao, M.T.W. Teo

Study supervision: A.E. Kiltie

The authors thank Ms. Filomena Estevez for her expert help with immunohistochemistry staining and Drs. Sameer Chilka for outlining the muscle-invasive tumor areas, Paul Manley of Novartis and Daniel Ridley of Glaxo Smith Kline for critical reading of the article.

This study was supported by Yorkshire Cancer Research project grant L336 and Cancer Research UK; Yorkshire Cancer Research Project Grant L336 (B. Qiao); Slovene Human Resources Scholarship Fund (B. Groselj); Yorkshire Cancer Research Project Grant L350 (M.T.W. Teo); Cancer Research UK Program grant C6228/A7675 (M.A. Knowles); Canadian Foundation for Innovation and Canadian Cancer Society and a Canadian Cancer Research Society Career Scientist Award (R.G. Bristow); and Cancer Research UK Program Grant C5255/A12678 (M. Kerr and A.E. Kiltie).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.