Targeting Homologous Recombination by Pharmacological Inhibitors Enhances the Killing Response of Glioblastoma Cells Treated with Alkylating Drugs

Glioblastoma multiforme is the most prevalent type of primary malignant brain tumor in adults, which is known for its aggressive progression, weak response to cancer therapy and, consequently, bad prognosis (1, 2). The poor overall 5-year survival rate of less than 10% points to the need of new chemotherapeutic approaches. The today's standard care of glioblastoma multiforme includes maximal surgical resection followed by radiation and concomitant temozolomide treatment followed by adjuvant temozolomide cycles (1–3). Chloroethylating nitrosoureas (CNU), including lomustine (CCNU), nimustine (ACNU), semustine (MeCCNU), carmustine (BCNU) and fotemustine, have been used as second-line drugs for the treatment of glioblastoma multiforme. Lomustine can be administered orally (4) and, therefore, it is currently preferentially in use. CNUs can also be applied for first-line therapy in the form of BCNU wafers (Gliadel) installed into the operative wound, the so-called localized chemotherapy for high-grade glioma (1). CNUs are also applied for second-line chemotherapy of metastatic melanoma, Hodgkin lymphoma, and brain metastases of different origin (5–10).

The mechanism of alkylating agents including CNUs has been intensively studied (11). They chloroethylate DNA and induce a broad spectrum of DNA adducts with O⁶-chloroethylguanine being the main killing lesion. This adduct is unstable and undergoes intramolecular rearrangement leading to an intermediate, N1-O⁶-ethanoguanine, and finally to N1-guanine-N3-cytosine interstrand cross-link (ICL; ref. 12). The critical primary lesion, O⁶-chloroethylguanine, is subject to repair by transfer of the alkyl group to O⁶-methylguanine-DNA methyltransferase (MGMT; ref. 13). This repair process renders cancer cells resistant to chloroethylating chemotherapeutics (14–17). The activity of MGMT in brain tumours is highly variable; 17% of pretreatment tumors do not display MGMT activity and up to 30% express MGMT at low level (18, 19). In these tumors, O⁶-chloroethylguanine remains unrepaired and the cytotoxic ICLs can be formed. In tumors with positive MGMT status, repeated treatment with a methylating anticancer drug causes MGMT to be used up by the repair reaction, which can be harnessed to deplete MGMT and to enhance the tumor cell killing response. Actually, coadministration of procarbazine and CCNU improved the therapeutic index for patients with glioblastoma multiforme (20).

ICLs are lesions that block replication fork movement. They can be removed by a complex repair pathway, allowing the restart of DNA replication (21). The repair of ICLs involves the transient formation of DNA double-strand breaks (DSB) by endonucleases at the site of the stalled replication fork (22, 23). Blocked replication and DSB intermediates activate the DNA damage response and downstream apoptotic pathways (24, 25). Since DSBs formed during ICL repair are critical secondary DNA lesions, it is reasonable to posit that their repair determines tumor cell resistance to CNUs. Two DSB repair pathways have been described, namely, homologous recombination (HR) and nonhomologous end-joining (NHEJ; ref. 26). NHEJ represents the major pathway for the repair of DSBs induced by ionizing radiation in G0 and G1 phase (27). It is considered to be error-prone, giving rise to chromosomal aberrations (28). In contrast, HR repair of DSBs is operative in late S–G2 and involves strand invasion onto the sister chromatid template, DNA repair synthesis, and resolution of the Holiday junction (29). Usage of the undamaged template ensures error-free DSB repair. Mre11 is an endo- and exonuclease, which participates in the recognition of DSBs via the MRN complex together with Rad50 and NBS1 (alias NBN). It also acts as a multifunctional DNA end-processing enzyme and plays a role in the restart of stalled replication forks by enhancing the resection at stalled forks, as shown for BRCA2-deficient cells (30, 31). The key HR protein is Rad51 because of its recombinase activity. Rad51 overexpression has been observed by immunohistochemistry in various cancers, for example, breast, pancreas, head and neck, lung, esophageal, and colon (32–36) and also in gliomas (37), whereas no detectable expression was found in normal differentiated tissues from the same organs (32–35). In most studies, Rad51 overexpression was associated with poor prognosis for the patients.

Investigating the role of HR and NHEJ in the cellular sensitivity to CNUs, we showed that HR-defective cells (BRCA2, Rad51D, and XRCC3 mutants) are hypersensitive to ACNU, whereas NHEJ-defective cells (Ku80 and DNA-PK mutants) were not or only mildly sensitive to the drug (25). This pertained the endpoints reproductive survival, apoptosis, and chromosomal breakage (25). Since HR plays a pivotal role in cellular protection against DNA alkylation–induced killing effects, it is reasonable to posit that pharmacologic inhibition of HR is a way of cancer sensitization. In the present study, we tested this hypothesis in human glioblastoma cell lines by targeting HR repair proteins pharmacologically. We used the MRE11 inhibitor mirin (38) and the Rad51 inhibitors B02 (39, 40) and RI-1 (41, 42) for inhibiting the DNA damage response (DDR) and HR, respectively, following treatment of cells with CCNU. We also studied RI-1 in combination with CCNU in a glioblastoma xenograft model. The data show that pharmacologic inhibition of MRE11 and HR is a feasible strategy for sensitizing glioblastoma cells to CNUs.T

The human glioma cell line LN229 was obtained from ATCC, U87MG from CLS Cell Line Service, and LN308 cells were provided by Dr. N. de Tribolet (Lausanne, Switzerland). LN229 and U87MG were authenticated by ATCC. LN308 was not further authenticated. The lines were kept frozen and used at early passage after receipt. They were previously characterized regarding their MGMT status (18, 43), karyotyped, and tested monthly for mycoplasma contamination. Cells were maintained in advanced DMEM (Life Technologies) supplemented with l-glutamine and 5% or, for the colony formation assay, 10% fetal calf serum (Gibco, Life technologies) at 37°C in a humidified atmosphere with 7% CO₂.

ACNU (nimustine hydrochloride from Sigma-Aldrich) was diluted in sterile bidistilled water, whereas CCNU (lomustine from Sigma-Aldrich) was diluted in absolute ethanol to give a concentration of 10 mmol/L and added to the cell culture medium or additionally diluted in medium to give final concentrations in the range of 5 to 50 μmol/L. In all assays, cells were pulse-treated with CCNU for 1 hour. Thereafter, the medium was replaced by fresh medium. Mirin (Tocris Bioscience), as well as B02 and RI-1 (both from Axon MedChem) were diluted in DMSO to give a concentration of 25 mmol/L and added to cell cultures to give final concentrations of 2.5 to 25 μmol/L. Cells were further incubated until harvesting. Chemical structures of CCNU, B02, mirin, and RI-1 are shown in Supplementary Fig. S1.

Cross-link formation and repair were measured by a modified version of the alkaline comet assay (single-cell gel electrophoresis, SCGE) as previously described (44). Briefly, exponentially growing LN229, LN308, or U87MG cells were treated for 1 hour with 50 μmol/L CCNU. Untreated and treated cells were harvested after 24 hours in PBS at a density of 2.5 × 10⁵ cells/mL and irradiated with 8-Gy γ-rays. The radiation was generated by ¹³⁷Cs (1,800 Ci) in a Gammacell 2000 device (Molsgaard Medical) at 7.7 rad/s (45). It is known that fragmentation of gDNA by ionizing radiation leads to comet formation, whereas induced ICLs reduce the mobility of DNA, respectively, the length and fluorescence intensity of the comets. Microgels were prepared by embedding single cells in low-melting-point agarose onto agarose-coated slides and standard alkaline SCGE was done as described (44). Propidium iodide (PI)-stained slides were evaluated using a fluorescence microscope (Nikon Instruments Europe) and the Comet IV software (Perceptive Imaging). The results are expressed as tail intensity comparing γ-ray irradiated cells after CCNU and cells treated with γ-rays only.

To measure the capacity of cells to repair DSBs by HR, LN229 cells were stably transfected with a pDRGFP plasmid (Addgene). The plasmid bears 2 nonfunctional GFP genes, one is truncated, the other containings a recognition site for I-SceI endonuclease. Upon transient transfection with I-SceI–expressing plasmid (Addgene), the endonuclease cleaves the modified GFP gene leading to a DSB. If this DSB is repaired via HR, by means of the sequence of the truncated GFP gene, a functional GFP protein is generated and can be quantified by flow cytometry. Cells were analyzed 72 hours after transfection with 1 μg pCβASceI using the Transfection Kit Effectene (Qiagen). During transfection, the cells were incubated in the presence or absence of 10 μmol/L B02, 25 μmol/L mirin, or 25 μmol/L RI-1. The DNA-PK inhibitor KU0060648 (Selleckchem) was used for comparison at a final concentration of 1 μmol/L. We expected that the inhibition of DNA-PK leads to a moderate increase in HR activity due to the lack of competition by NHEJ for the repair of DSB (46). Cells were trypsinized and washed with PBS and measured by flow cytometry using FACS Canto II (BD Biosciences). Data were analyzed with BD FACSDiva software.

The DNA fiber assay was performed as described (47). Briefly, exponentially growing cells were treated with 30 μmol/L CCNU and further incubated in the presence or absence of 10 μmol/L B02, 25 μmol/L mirin, or 25 μmol/L RI-1 for 6 or 16 h and then pulse-labeled with 25 μmol/L 5-chloro-2′-deoxyuridine (CldU; Sigma-Aldrich) followed by labeling with 250 μmol/L 5-iodo-2′-deoxyuridine (IdU; TCI Deutschland) for 30 minutes each. Labeled cells were harvested and DNA fiber spreads prepared. Acid-treated fiber spreads were stained with monoclonal rat anti-BrdUrd (Oxford Biotechnologies, 1:1,000) followed by monoclonal mouse anti-BrdUrd (Becton Dickinson, 1:1,500). Primary antibodies were detected by goat anti-rat Fab2 Cy3–coupled (Jackson ImmunoResearch) and goat anti-mouse Fab2 Alexa488–coupled secondary antibodies (Life technologies, 1:500). Fibers were examined and images captured using LSM 710 with ZEN 2009 software (Zeiss). CldU (red) and IdU (green) tracks were measured using LSM Image Browser (Zeiss) and micromolar values were converted into kilo basepairs. At least 150 forks were analyzed from 3 repetitions. DNA fiber structures from 3 independent experiments were counted in ImageJ using the Cell Counter function.

In colony formation assays, glioblastoma cells growing in the log phase were used. Cells (n = 400 for LN229 and LN308, n = 800 for U87) were seeded in duplicate in 60-mm petri dishes. They were allowed to attach, exposed to 10 μmol/L CCNU for 1 hour, and then the medium was changed with or without addition of inhibitor. Cells were further incubated. After 7 to 10 days, the colonies were fixed with 70% ethanol for 10 minutes, stained in crystal violet solution (1 g/L dH₂O) for 10 minutes and colonies containing at least 40 cells were counted and presented graphically as a percentage of untreated cells (control). All colony assays were repeated at least 3 times.

To distinguish between early stages of apoptosis and late apoptosis/necrosis, we used Annexin V/PI (AV/PI) double staining of nonfixed cells as previously described (25). Briefly, the culture medium of untreated or treated cells was collected and then the cells were washed in PBS and detached by trypsin/EDTA, added to the culture medium and centrifuged. The cell pellets were washed with PBS and suspended in 50 μL Annexin-binding buffer (Miltenyi Biotec). Annexin V-FITC from Miltenyi Biotec (2.5 μL) was added to each sample. After 15-minute incubation on ice in the dark, 430 μL binding buffer and 1 μg/mL PI were added. The flow cytometric measurement was carried out using FACS Canto II (BD Biosciences). For each sample, 10,000 cells were scored. The percentage of the total induced cell death and the ratio between apoptosis and necrosis in mock-treated and CCNU inhibitor–treated samples, respectively, were determined by using BD FACS Diva software (BD Bioscience). All results represent means of at least 3 independent experiments ± SD.

Additionally to cell death determination, we analyzed caspase activation via flow cytometry as previously described (48) with slight modifications. To this end, we used the cell-permeable fluoromethyl ketone (FMK)-derived peptide zVAD-FMK, which binds covalently to the catalytic site of caspase proteases and acts as an effective irreversible general caspase inhibitor without additional cytotoxic effects. Briefly, harvested cells were resuspended in 250 μL PBS and incubated for 60 minutes in the presence of 20 μmol/L of fluorescein isothiocyanate (FITC)-coupled-zVAD-FMK (CaspACE FITC-VAD-FMK, Promega) in the incubator. Following incubation, cells were rinsed twice in cold PBS, fixed with ice-cold 70% ethanol, and stored at −20°C for at least 2 hours. Prior to flow cytometry, cells were digested in RNAse A (0.03 mg/mL PBS, 1 hour, room temperature), counterstained with PI (16.5 μg/mL), and stored on ice until measurement. A total of 20,000 events were acquired and analyzed with BD FACS Diva software (BD Bioscience). The percentage of cells in the different phases of the cell cycle or in the sub-G1 fraction and the percentage of cells with activated caspases (FITC-positive cell subpopulation) was determined by the same software. The results represent means of at least 3 independent experiments ± SD.

Cells grown on precleaned cover slips were treated with CCNU and/or inhibitors. Following posttreatment incubation, the cells on the cover slips were fixed with ice-cold methanol/acetone (7:3, v/v) for 10 minutes at −20°C at the indicated time points. Following rehydration in PBS, the fixed cells were blocked/permeabilized in 10% normal goat serum + 0.25% Triton X-100 in PBS for 1 hour. Furthermore, the preparations were incubated with primary antibodies: mouse anti-phospho-H2AX (Ser139), anti-RPA2 from Merck Millipore (Germany), and rabbit monoclonal anti-phospho H2AX (Ser139) from Cell Signaling Technology. We used secondary anti-mouse and anti-rabbit antibodies coupled with Alexa Fluor 488 (Life technologies) or Cy3 (Jackson ImmunoResearch Europe). Nuclei were counterstained with DAPI containing Vectashield mounting medium (Vector Labs) for Metafer and To-Pro-3 for LSM microscopy. The slide-scanning platform Metafer (MetaSystems) was used with Metafer4 software for automatic capture of images. Scoring of foci on captured images was performed with ImageJ software with suitable batch-macros as described (49). Representative confocal images were captured by ZEN2009 software for laser scanning microscope LSM710 (Carl Zeiss). EdU (5-ethynyl-2′-deoxyuridine) incorporation for detection of S-phase cells was performed using the EdU Click-It Imaging Kit (Thermo Fisher Scientific) and FAM-azide click-it dye (Lumiprobe). The intensity of the Click-It fluorescent dye bound to EdU in the DNA of EdU-labeled cells was measured onto LSM images by ZEN 2009 Software. EU (5-ethynyl-2′-uridine) incorporation for analysis of transcription rate was performed using the Click-iT RNA Alexa Fluor 488 Imaging Kit (Thermo Fisher Scientific). The intensity of the Click-It fluorescent dye bound to EU was measured for each cell captured onto LSM images using the ImageJ software.

Protein extracts were prepared from LN229 cells following 1-hour treatment with 30 μmol/L CCNU and posttreatment with 10 μmol/L B02, 25 μmol/L mirin, or 20 μmol/L RI-1. Cells were detached from the dishes by trypsin/EDTA treatment and collected by centrifugation. The cell pellets were lysed, sonified and boiled in loading buffer at 95°C for 5 minutes, and then cooled on ice for 5 minutes. Per slot, 30 μg protein was loaded onto 5% SDS stacking gel/7.5%, 10%, or 15% separation gel (19:1, stock polyacrylamide: bisacrylamide) and run at 40 mA until the loading buffer indicator leaves the gel. Proteins were blotted onto a nitrocellulose membrane. Membranes were blocked for 1 hour at room temperature in 5% (wt/v) albumin fraction of bovine serum in TBS containing 0.1% Tween (TBST), then washed 2 times in TBST, and then incubated overnight at 4°C with the primary antibody, that is, anti-H2AX (phospho-Ser139, 1:10,000) from Millipore-Merck; rabbit monoclonal anti-cleaved PARP (1:100,000) from Abcam; rabbit anti-p53 (phospho-Ser15), anti-H2AX (phospho-Ser139), and rabbit monoclonal anti-PTEN (1:1,000) from Cell Signaling. After washing in TBST, the membranes were incubated with a donkey IRDye 800CW or IRDye 680 anti-mouse or anti-rabbit secondary antibody (1:10,000–1:25,000 from LI-COR Biosciences), dried and imaged by Odyssey v3.0 of the infrared imaging device Odyssey (LI-COR Biosciences). The mouse anti-β-actin (Santa Cruz) was used as a loading control (1:3,000) followed by a secondary anti-mouse antibody.

Cells grown in 4-mL medium in 60-mm petri dishes were treated with 15 μmol/L CCNU for 60 minutes 1 cell cycle prior to fixation because we expected chromosome aberration induction during the S-phase of the treatment cell cycle. Demecolcine (150 ng/mL) was added 12 hours before harvest. Chromosome preparations were done according to standard protocols. Briefly, cells were harvested by trypsin:EDTA treatment, pelleted, and resuspended in prewarmed 0.075 mol/L KCl and incubated for 10 minutes. Cells were pelleted and fixed 3 times in ice-cold methanol/acetic acid (3:1, v/v) mixture. The fixed cells were resuspended in a small volume of freshly prepared ice-cold fixative and dropped onto precleaned wet ice-cold slides. After being air-dried, the slides were stained in 5% phosphate-buffered Giemsa solution. Fifty metaphases were evaluated per treatment level for chromosomal aberrations. The following aberrations were counted: chromatid breaks, chromatid translocations (triradials, quadriradials), and intercalary deletions. Gaps were not included in the final evaluation. Because of the variable numbers of chromosomes in the different cell lines used in this study, aberration frequencies were expressed as aberrations per chromosome and aberrations per metaphase (normalized to mean number of chromosomes for 50 metaphases of each treatment variant). Induced aberration frequencies for combined treatments in 100 metaphases were calculated by subtraction of the aberration frequency induced by HRi alone and compared with CCNU treatment alone.

The animal experiments were approved by the government of Rhineland-Palatinate and the Animal Care and Use Committee of the University Medical Center Mainz and performed according to the federal law (Animal Protection Act of Germany). Immunodeficient mice (Balb c, nu/nu, Janvier Labs) were housed in a sterile environment and allowed free access to food and water. Animals were 3 to 4 months of age at the start of the experiment. Tumor cells were injected subcutaneously over both flanks with 2.5 × 10⁶ U87MG cells per flank. The animals were checked every other day to evaluate their health condition and tumor growth. When the tumors reached an average diameter of about 5 mm, the animals were divided into groups of 5 animals that were subjected to different treatments: to the animals of first group (vehicle control), corn oil (Sigma-Aldrich) alone was administered intraperitoneally; second group, CCNU (25 μg/kg body weight diluted in 50 μL corn oil) was injected intraperitoneally, the third group was treated intraperitoneally with 50 mg/kg RI-1 diluted in 50 μL corn oil, and the fourth group with the combination CCNU + RI-1. The health state, body weight, and tumor growth were assessed twice a week, and tumor volumes were calculated using the formula V_t = l × b × 0.5b, where V_t is the tumor volume, l is the length, and b the width of the tumour.

We used the software GraphPad Prism Version 6.00 for Windows by GraphPad Software for statistical analyses. Datasets from 3 independent experiments were compared by one-tailed unpaired t test or by unpaired t test with Welch correction (if variances between the datasets were significantly different). For chromosomal aberrations, the data on CCNU-treated cells with/without posttreatment with HRi from 3 independent experiments were compared by the nonparametric Mann–Whitney test. Tumor growth curves were compared by the Mann–Whitney test as well.

We used 3 well-established glioblastoma cell lines deficient in MGMT activity as an in vitro model (50). The cell lines differ in their p53 status: LN229 and U87MG are p53 wild-type and LN308 is p53-mutated. Consequently, following treatment with ACNU or CCNU, p53 was phosphorylated (at serine 15) in LN229 and U87MG, whereas LN308 did not show phosphorylation of p53 (Fig. 1A). The lines differ also in their PTEN status; LN229 is PTEN wild-type, while LN308 and U87MG are PTEN-deficient (Fig. 1A). Using a modified alkaline comet assay for the detection of crosslinks, we observed that CCNU treatment leads to the formation of ICLs in glioblastoma cells. Thus, as shown in Fig. 1B, the tail intensity was significantly reduced in the irradiated cell lines pretreated with 50 μmol/L CCNU, indicating the induction of ICLs.

To prove that the repair inhibitors B02, mirin, and RI-1 are effective in the glioblastoma lines used here, we applied a plasmid-based repair assay. In cells transfected with pDRGFP, a DSB was generated in the plasmid sequence by I-SceI, and the repair of the DSB by HR leads to GFP expression. As shown in Fig. 1C, the HR capacity observed in the control was significantly reduced following B02, mirin, and RI-1 treatment. In contrast, DNA-PK inhibition caused a strong increase in the number of GFP positive-cells, probably due to a compensatory enhancement of HR activity in the absence of the competitive NHEJ DSB repair pathway (46).

Furthermore, we investigated whether O⁶ClG-mediated ICLs, which were induced at the dose levels applied in the subsequent inhibitor experiments, block DNA replication. To this end, we analyzed the incorporation of the thymidine analogue EdU in the DNA following CCNU treatment. Representative images of EdU incorporation are shown in Fig. 2A. The quantification shows that 6 hours after pulse treatment (60 minutes) with CCNU, the EdU incorporation was not significantly reduced (Fig. 2B), whereas 16 hours later, a significant reduction was observed (Fig. 2C). The HR inhibitors themselves had a replication-inhibiting effect 16 hours after treatment. The inhibitors B02 and mirin given together with CCNU had no significant impact on EdU incorporation as measured 6 hours after CCNU, but slightly ameliorated the effect 16 hours later, whereas RI-1 did not (Fig. 2C).

To gain insight into the mechanism of replication inhibition, we made use of the DNA fiber labeling technique, where the DNA replication tracks are visualized through the detection of incorporated thymidine analogues. The distribution of the DNA fiber structures (defined in Fig. 2D together with representative images) showed an increase in the frequency of stalled replication forks (replication interrupted during the first CldU pulse; red segments in the columns) at 6 and 16 hours after CCNU treatment, with an exacerbated effect 16 hours after treatment. The HR inhibitors alone had an impact on ongoing replication, which was ameliorated if applied together with CCNU. Interestingly, RI-1 abrogated the effect of CCNU if measured 16 hours after treatment (Fig. 2E and F). Overall, the data support the notion that at the dose level used here (resulting in moderate toxicity) CCNU-induced DNA lesions, very likely ICLs, only modestly inhibit replication and that the inhibitors B02 and mirin, but not RI-1, ameliorate the replication-blocking effect of CCNU.

CCNU is a potent inducer of replication protein A (RPA) foci (see Fig. 3A for representative images), which are considered as a marker of replication stress (51). The quantitative analysis showed that high levels of RPA foci were induced in LN229 cells 24 hours after pulse treatment with 30 μmol/L CCNU (Fig. 3B). After 72 hours, the amount of RPA foci declined to nearly control level, indicating that damage repair has occurred. Similar numbers of RPA foci/cell were induced 24 hours after combined treatments with CCNU plus either one of the inhibitors; however, no significant decline was observed at 72 hours (Fig. 3B). This indicates that RPA-inducing lesions are formed but not repaired if cells are treated with CCNU followed by mirin or an inhibitor of HR.

Next, we analyzed DSBs formed following CCNU pulse treatment with and without post exposure to the HR inhibitors RI-1 and B02 and the MRE11 inhibitor mirin. In Fig. 4A, representative examples are shown of γH2AX foci induced in LN229 cells 16 hours after treatment (for examples of 6-hour measure points, see Supplementary Fig. S2) and Fig. 4B presents their quantification. The data demonstrate that a low number of DSBs are formed 6 hours after CCNU treatment while 16 hours later DSBs were significantly accumulated. The presence of the inhibitors in the postincubation period clearly enhanced the DSB frequency at both time points. The data suggest that inhibition of HR and MRE11 ameliorates the DSB level.

In another experimental series, γH2AX foci were determined after longer times, that is, 24 and 72 hours after CCNU pulse treatment, comparing the cell lines. Representative images for LN229, LN308, and U87MG are shown in Fig. 5A. The quantification revealed that high γH2AX foci levels were induced 24 hours after CCNU pulse treatment or the combined treatments with the inhibitors. At 72 hours, the cells repaired most of the CCNU-induced DSBs. However, in the presence of the inhibitors, the γH2AX foci level did not decline, indicating that DSBs induced by CCNU were not repaired (Fig. 5B–D). To confirm the data, we analyzed the level of phosphorylated γH2AX in Western blots. The experiments showed an increase in γH2AX protein level after CCNU treatment alone. Following treatment with CCNU together with B02, mirin, or RI-1, the initial level was the same (24 hours after the onset of treatment), but the levels 48 and 72 hours after treatment were clearly higher in the combined treatments than in the control CCNU only. The strongest effect was observed for B02 applied together with CCNU (Fig. 5E). The finding indicates that in the presence of B02, mirin, and RI-1 DSBs induced by CCNU remained unrepaired, at least for a period of 72 hours. In summary, the data shows that DSBs are formed in MGMT lacking glioblastoma cells treated with CCNU. These DSB are subject to repair. Inhibition of HR by B02 or RI-1 or targeting of MRE11 by mirin exerted an inhibitory effect on the repair of DSBs formed following CCNU treatment.

Next, we analyzed the formation of CCNU-induced chromosomal aberrations. The 3 glioblastoma cell lines used in our study differed in their karyotype: LN229 and LN308 were hypertriploid with about 82 and 75 chromosomes per metaphase, respectively, whereas U87MG cells were aneuploid (44–47 chromosomes per metaphase). We could show a significant increase of CCNU-induced structural chromosomal aberration after Rad51 inhibition with RI-1 and with B02 in all cell lines (Fig. 6A–C). For mirin, we observed a sensitizing effect only in LN308 (Fig. 6B). Under the same conditions (using the same concentrations of CCNU and HR inhibitors), the U87MG cells were much more resistant than LN229 and LN308 cells. The analysis of the aberration spectra revealed the predominant formation of chromatid-type aberrations, that is, breaks and exchanges. The LN229 cells were very sensitive to Rad51 inhibition and especially after treatment with the combination of B02 and CCNU, many multiaberrant metaphases were observed with more than 50% of the chromosomes involved in complex rearrangements (see representative image in Fig. 6A). These data demonstrate the clastogenic effect of CCNU in glioblastoma cells, which is exacerbated if posttreatment occurred with either one of the inhibitors. In summary, we show that inhibition of Rad51 and MRE11 ameliorates the genotoxic effects of CCNU in glioblastoma cells.

To explore the effect of HR and MRE11 inhibitors on reproductive cell survival, we conducted colony-forming assays. The data are shown in Fig. 7. We observed a significantly reduced survival after CCNU followed by B02, mirin, or RI-1 in LN229 cells (Fig. 7A), following B02 and RI-1 in LN308 cells (Fig. 7B) and following RI-1 in U87MG cells (Fig. 7C). Thus, the most consistent results as to enhancing reproductive cell death followed by CCNU were obtained for RI-1.

To further study the effect of the inhibitors on glioblastoma cell sensitivity, we used the AV/PI assay, measuring apoptosis and necrosis. The data shown in Fig. 7D–F demonstrate that CCNU-induced cell death by apoptosis and necrosis is ameliorated when CCNU was applied concomitantly with RI-1, mirin, and B02 (Fig. 7D for LN229, Fig. 7E for LN308, and Fig. 7F for U87MG). Although both apoptosis and necrosis were induced, apoptosis was prevailing after CCNU treatment alone and after the combined treatments. The increased levels of apoptosis following treatment with the inhibitors were confirmed by caspase assays, showing a higher percentage of cells with activated caspases following treatment with CCNU plus HRi (Fig. 8A–C for LN229, LN308, and U87MG, respectively). Apoptosis induced by CCNU was accompanied by PARP-1 cleavage (Fig. 8D), which was clearly higher after Rad51 inhibition by B02 and RI-1 in LN229 cells. Mirin alone induced already PARP1 cleavage, which is in line with its toxicity seen in other assays.

Having shown that RI-1 is able to enhance CCNU-induced glioma cell death in a nearly nontoxic dose range in vitro, we investigated the toxicity of RI-1 in vivo alone and in combination with CCNU in a nude mouse xenograft model. Administration of RI-1 did not result in changes in the health status or body weight during a 3-week postexposure period (data not shown). Representative images of U87MG tumor xenografts are shown in Fig. 9A, demonstrating the impressive effect of combination treatment CCNU plus RI-1. The tumor growth was quantified, revealing that CCNU treatment leads to a reduction in tumor size which is greatly ameliorated if CCNU is combined with RI-1 (Fig. 9B). Figure 9C shows body weight of the mice in the different treatment groups, which was not clearly impaired in the combination treatment setting. The data let us conclude that the combined treatment CCNU plus RI-1 is tolerable and has an advantage over CCNU alone as to inhibition of tumor growth in the xenograft model.

This study was aimed at analyzing whether inhibition of HR or inhibition of MRE11 results in an enhancement of the killing effect of CCNU, which is a representative of CNUs used in glioblastoma therapy. We show that the inhibitors RI-1 and B02, which target Rad51, enhance the level of DSBs and chromosomal aberrations and ameliorate the killing effect of CCNU in glioblastoma cells. The same was observed for mirin, which targets MRE11, a component of the MRN complex.

CCNU, like other chloroethylating agents, alkylates DNA at several sites (52), including the O⁶-position of guanine. Intramolecular rearrangement of O⁶-chloroethylguanine gives rise to ICLs between guanine and cytosine. Since MGMT, which repairs O⁶-chloroethylguanine, nearly completely abolishes the cytotoxicity of CNUs, at least in the pharmacologically low-dose range (25), it can be extrapolated that this minor damage, O⁶-chloroethylguanine, and the subsequently formed ICLs are responsible for the genotoxicity and the killing effect of CNUs. ICLs block DNA replication, which was observed in our experiments within the cytotoxic dose range measuring EdU incorporation at times when ICLs were formed (16 hours following treatment). The inhibitory effect on EdU incorporation was, however, quite modest. This is likely the result of the small amount of ICLs induced in comparison to monoadducts. In the DNA fiber–labeling experiments, we observed already 6 hours after CCNU pulse treatment an increase in the frequency of uniformly labeled tracks (only CldU or IdU incorporated), which likely represent blocked replication forks during the labeling period. Since blockage of fork movement results in stalled replication forks, replication fork collapse and finally DSBs according to a mechanism described for other genotoxic insults like UV or MMS (53, 54), we posit that the same mechanism takes place following CCNU-induced toxic DNA adducts, very likely ICLs. The significant formation of DSBs at this early time point and their further accumulation support this notion.

DSBs formed by alkylating agents are subject to repair by HR, as previously shown for both O⁶-methylguanine–inducing anticancer drugs (55, 56) and chloroethylating agents (25, 55, 57). Therefore, it was reasonable to hypothesize that inhibition of HR ameliorates DSB formation. The data presented here can be taken to confirm that this is indeed the case. Whereas without HR inhibition, the level of γH2AX declined 72 hours after formation, they remained at a high level in RI-1 and B02 post-treated cells. This was associated with a significant increase in chromosomal changes and killing effects, notably by the induction of apoptosis. RI-1 binds directly to human RAD51 at cysteine-319 (42). For B02, a specific binding to the Rad51 protein was also demonstrated, but the exact binding site was not identified (39). In each case, the binding of Rad51 to single-stranded DNA becomes more difficult, which inflicts on the HR process. Comparing the inhibitors, we observed that RI-1 is less cytotoxic than B02 in glioblastoma cells and, overall, slightly more efficient in enhancing CCNU-induced genotoxicity and cell death.

Another goal of this work was to compare pharmacologic inhibitors of HR with an inhibitor of DNA damage signaling, mirin. The MRN complex plays a profound role in the recognition of DSBs. MRE11 is a key protein in the trimeric MRN complex, composed of NBS1, RAD50, and MRE11. MRE11 contains a core phosphodiesterase domain that is responsible for its multiple nuclease activities in a single endo/exonuclease mechanism (58). Mirin was shown to bind to MRE11 and to inhibit its nuclease activity without affecting MRE11 binding to DNA or the MRN-associated DNA tethering activity (38). In addition, it inhibits the ATM-dependent phosphorylation of NBS1 and CHK2 (38). Therefore, the activation of the ATM/ATR-CHK1-CHK2 pathway cannot be triggered in a proper way. Since MRN and ATM operate independently in the recognition of DNA damage and processing of DSBs for HR repair (59), it was hypothesized that the pharmacologic inhibition of the MRN complex exerts a sensitizing effect similar to what we observed with Rad51 inhibitors. This was indeed the case, as mirin ameliorated the level of DSBs, aberrations, and apoptosis/necrosis in CCNU-treated glioblastoma cells similar to Rad51 inhibitors. We should note that a combined treatment with mirin and Rad51 inhibitor was heavily toxic for the cells.

Previously, it was shown that downregulation of Rad51 or BRCA2 by means of siRNA enhances the killing response of glioblastoma cells treated with temozolomide or CCNU (55, 57). On principle, virus-mediated transducing systems might represent a reasonable approach for targeting HR in glioblastoma therapy. However, it is clear that viral transduction strategies are more difficult in finding their way to the clinique than pharmacologic inhibitors. Therefore, we favor the use of HR inhibitors as a strategy for enhancing the effect of alkylating drugs in glioma therapy. When comparing the inhibitors used in this study, we observed for B02 and for mirin an inhibiting effect on transcription, which might impact the survival of non-cancer cells. RI-1 on the other hand did not affect transcription (Supplementary Fig. S3) and, therefore, might have less side effects on nontarget tissues. Actually, RI-1 was well-tolerated in the in vivo setting and potent in ameliorating the cytotoxic and antitumor activities of CCNU, as shown in the U87MG xenograft model. It is clear that clinical safety trials are needed for assessing the systemic toxicity of the drugs. Although the RAD51 inhibitors have been around for a number of years (60), they have not yet moved to extended clinical trials very likely because supporting preclinical data were lacking. This study hopefully encourages the application of HR inhibitors in trials assessing their effect in combination with alkylating drugs that are used in the therapy of glioblastomas and other tumor groups.

No potential conflicts of interest were disclosed.

Conception and design: B. Kaina, T. Nikolova

Development of methodology: N. Berte, K. Borgmann, T. Nikolova

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Berte, N. Piecha, M. Wang, T. Nikolova

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Berte, A. Piee-Staffa, N. Piecha, M. Wang, K. Borgmann, B. Kaina, T. Nikolova

Writing, review, and/or revision of the manuscript: N. Berte, K. Borgmann, B. Kaina, T. Nikolova

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Piee-Staffa, B. Kaina, T. Nikolova

Study supervision: T. Nikolova

The authors thank Anna Frumkina and Georg Nagel for technical assistance, and are very grateful to Ann Parplys for critical reading and helpful comments on this article.

This work was supported by DFG (German Research Foundation), grant Ni1319/1–1 and Ni1319/1-2, and by a grant from the Internal Research Funding of University Mainz to T. Niklova (2015–2016).

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.