Glioblastoma (GBM) is the most common and aggressive primary brain tumor in adults and is universally fatal. The DNA alkylating agent temozolomide is part of the standard-of-care for GBM. However, these tumors eventually develop therapy-driven resistance and inevitably recur. While loss of mismatch repair (MMR) and re-expression of MGMT have been shown to underlie chemoresistance in a fraction of GBMs, resistance mechanisms operating in the remaining GBMs are not well understood. To better understand the molecular basis for therapy-driven temozolomide resistance, mice bearing orthotopic GBM xenografts were subjected to protracted temozolomide treatment, and cell lines were generated from the primary (untreated) and recurrent (temozolomide-treated) tumors. As expected, the cells derived from primary tumors were sensitive to temozolomide, whereas the cells from the recurrent tumors were significantly resistant to the drug. Importantly, the acquired resistance to temozolomide in the recurrent lines was not driven by re-expression of MGMT or loss of MMR but was due to accelerated repair of temozolomide-induced DNA double-strand breaks (DSB). Temozolomide induces DNA replication–associated DSBs that are primarily repaired by the homologous recombination (HR) pathway. Augmented HR appears to underpin temozolomide resistance in the recurrent lines, as these cells were cross-resistant to other agents that induced replication-associated DSBs, exhibited faster resolution of damage-induced Rad51 foci, and displayed higher levels of sister chromatid exchanges (SCE). Furthermore, in light of recent studies demonstrating that CDK1 and CDK2 promote HR, it was found that CDK1/2 inhibitors countered the heightened HR in recurrent tumors and sensitized these therapy-resistant tumor cells to temozolomide.
Implications: Augmented HR repair is a novel mechanism underlying acquired temozolomide resistance in GBM, and this raises the possibility of improving the therapeutic response to temozolomide by targeting HR with small-molecule inhibitors of CDK1/2. Mol Cancer Res; 14(10); 928–40. ©2016 AACR.
Glioblastomas are incurable, malignant brain tumors (1). The median survival of patients with glioblastomas is only slightly above a year, and 5-year survival rate is less than 10% despite aggressive treatment regimens (2). The current mainstay of treatment is surgical resection, followed by radiation therapy combined with concomitant and adjuvant temozolomide chemotherapy. Addition of temozolomide to the standard therapy increases the survival rates significantly albeit minimally (2, 3). This lack of response to temozolomide therapy is mostly due to intrinsic and acquired resistance of tumors to the drug (4–12). With very limited alternative chemotherapeutic agents for glioblastomas (13), temozolomide-resistant patients are virtually left without treatment options after tumor recurrence. Therefore, it is crucial to understand mechanisms of resistance to develop strategies to resensitize these tumors to temozolomide.
Temozolomide is an alkylating agent that methylates DNA at the N7position of guanine, the O3 position of adenine, and the O6 position of guanine (14). The cytotoxicity of temozolomide is mostly due to O6-methylguanine (O6-meG; refs. 15–17). If not directly reversed by O6-methylguanine methyl transferase (MGMT), O6-meG mispairs with thymine during DNA replication. The mismatch is recognized by the mismatch repair (MMR) machinery, which promotes excision of the newly synthesized strand leaving the parental strand with the O6-meG lesion intact (14). O6-meG can then mispair again with thymine leading to repeated cycles of MMR attempts and persistent single-stranded gaps in DNA (18). During the next round of replication, replication forks encounter single-stranded DNA, and one-ended DNA double-strand breaks (DSB) are formed (19, 20). Such replication fork–associated DSBs can be correctly repaired only by homologous recombination (HR), an error-free DSB repair pathway which operates in the S and G2 phases of the cell cycle (21, 22). If left unrepaired or if repaired by the error-prone nonhomologous end joining (NHEJ) pathway (23), such breaks would be extremely toxic to the cell (24).
Although temozolomide-induced DSBs are very toxic, the efficacy of temozolomide is greatly limited by both inherent and acquired resistance mechanisms that prevent the induction of such breaks. MGMT can directly reverse the O6-meG lesions; therefore, high expression of the enzyme leads to temozolomide resistance presumably by preventing the generation of DSBs. Conversely, MGMT promoter methylation (which silences expression of MGMT) correlates with a better response to temozolomide (25, 26). Thus, acquired resistance has been correlated with increased expression of MGMT in recurrent tumors in clinical and mouse xenograft studies (4–7). Another reported mechanism of acquired resistance is loss of MMR (4, 6–12), the basic principle being that without MMR, single-stranded DNA stretches are not generated, and consequently, no DSBs are formed at replication forks.
Although important, MGMT expression and MMR loss do not account for resistance to temozolomide in all cases (7, 27). On the basis of MGMT promoter methylation status, presumably only about half of GBMs express MGMT (25, 28), and only a fraction display evidence of MMR deficiency (6, 8, 27), indicating that there must exist yet undiscovered mechanisms of resistance. Here, we report that protracted treatment of MGMT-deficient orthotopic GBM tumors with temozolomide results in acquired resistance due to accelerated repair of temozolomide-induced DSBs. Moreover, we show that this resistance can be attributed to augmented homologous recombination (HR) and that dampening HR in such tumors sensitizes cancer cells to temozolomide therapy.
Materials and Methods
Cell culture and temozolomide treatment in vitro
GBM9 neurosphere cultures were maintained as described previously (29, 30). Briefly, neurospheres were maintained in DMEM/F12 1:1 media (Life Technologies) supplemented with B27 without vitamin A (Life Technologies), 10 ng/mL EGF (PeproTech), and 10 ng/mL bFGF (PeproTech). Human glioma cell lines (T98G, U138, and U251) were maintained in DMEM (Corning) with 10% FBS (Hyclone) and 1% penicillin/streptomycin (Gibco). Cell culture media for U138 were supplemented with nonessential amino acids (Gibco). All cells were maintained in a humidified incubator at 37°C with 5% CO2. Glioma cell lines were purchased from ATCC and grown for less than 6 months after revival. All cells were tested to be Mycoplasma free. To generate temozolomide-resistant lines in vitro, GBM9 neurospheres, and T98G, U138, and U251 monolayer cultures were treated with 50 μmol/L temozolomide, which was replenished every other day, over 24 days. Temozolomide (Sigma-Aldrich) was dissolved in DMSO (Sigma-Aldrich) to generate a 100 mmol/L stock which was stored at −80°C.
Animal injections, treatments, and ex vivo culture generation
Animal studies were performed in accordance with UT Southwestern IACUC–approved protocols. Intracranial tumors were generated by injecting 5 × 105 GBM9 cells in 5 μL of growth media into the right corpus striatum of nu/nu nude mice (Charles River, Stock#88), as described previously (30). Treatment was initiated 7 days after injection. Mice were treated with temozolomide (20 mg/kg) by oral gavage (vehicle: polyethylene glycol 300; Sigma-Aldrich) or with vehicle only as a control; 12 doses were administered every other day over a 24-day period. Mice bearing intracranial tumors were sacrificed when they became moribund. Brains were removed and tumors dissected out under a dissecting light microscope. Ex vivo cultures were generated by triturating tumor tissue in trypsin (Sigma-Aldrich) and culturing the triturate in plastic flasks in DMEM (Corning) supplemented with 10% FBS (Hyclone) and 1% penicillin/streptomycin (Gibco).
Cells were plated in triplicate onto 60-mm dishes (1,000 cells per dish) and treated with ionizing radiation (cesium source; J.L. Shepherd and associates), temozolomide (Sigma-Aldrich), N-methyl-N'-nitro-N-nitrosoguanidine (MNNG; Sigma-Aldrich), camptothecin (Sigma-Aldrich), or etoposide (Sigma-Aldrich) at the indicated doses. MNNG, camptothecin, and etoposide were dissolved in DMSO (Sigma-Aldrich) to generate 10 mmol/L stocks which were stored at −80°C. At 48 hours after temozolomide or MNNG, 2 hours after camptothecin, or 1 hour after etoposide addition, drug-containing medium was replaced with drug-free medium. For chemosensitization experiments, temozolomide-containing medium was replaced with medium containing either 5 μmol/L roscovitine (Sigma-Aldrich) or 0.25 μmol/L AZD5438 (Selleck) for 48 hours, after which drug-free medium was added to allow colony formation. Roscovitine and AZD5438 were dissolved in DMSO (Sigma-Aldrich) to generate 10 mmol/L stocks which were stored at −80°C. Surviving colonies were stained with crystal violet 10 to 14 days later, as described before (31). For the sphere formation assay, neurospheres were dissociated into single cells and serially diluted in temozolomide-containing medium with the indicated drug concentrations. Cells were transferred to 96-well plates at a concentration of 5 cells per well. Spheres were allowed to grow for 14 days, and total numbers of neurospheres was quantified visually using a light microscope, as described before (30).
Whole-cell extracts were prepared and Western blotted as described before (31). Antibodies used were as follows: actin (Sigma-Aldrich); CHK1 (Cell Signaling); phospho-CHK1 (Ser317; Cell Signaling); CHK2 (Cell Signaling); phospho-CHK2 (Thr68; Cell Signaling); MGMT (Millipore); MLH1 (BD Biosciences); MSH2 (BD Biosciences); MSH3 (Assay Biotech); MSH6 (Assay Biotech); and horseradish peroxidase (HRP)-conjugated secondary antibodies (Biorad).
Cells were treated with 10 μmol/L temozolomide for 48 hours or with 10 Gy of ionizing radiation (cesium source; J.L. Shepherd and associates) and fixed at the specified time points in ethanol overnight at −20°C. Cell-cycle stage was assessed by single-parameter flow cytometry (after propidium iodide staining for DNA content) using a BD CYTOMICS FC500 Flow Cytometer (Becton Dickinson), as described before (32).
gDNA was obtained from high-passage (passage 12 or greater) ex vivo cultures by using the DNeasy Blood and Tissue kit (Qiagen). Microsatellite-containing loci were amplified by PCR and products electrophoresed in 8% PAGE to determine their relative sizes. PCR reaction conditions and primers were used as described (33).
DNA DSB repair
For monitoring DSB repair, cells were seeded in chamber slides (Lab-Tek) and treated with temozolomide (10 μmol/L) for 48 hours, MNNG (1 μmol/L) for 48 hours, camptothecin (50 nmol/L) for 2 hours, or etoposide (2 μmol/L) for 1 hour, or irradiated at 1 Gy (cesium source; J.L. Shepherd and associates). For assessing the effects of CDK inhibition on DSB repair, cells were first treated with temozolomide (10 μmol/L) for 48 hours after which temozolomide-containing medium was replaced with medium containing either 5 μmol/L roscovitine (Sigma-Aldrich) or 0.25 μmol/L AZD5438 (Selleck). Cells were fixed at the indicated times and processed for immunofluorescent staining with anti-53BP1 antibody (Santa Cruz), as described before (31). The average numbers of 53BP1 foci per nucleus were determined after scoring at least 50 nuclei and subtracting background. Images were captured using a Leica DH5500B fluorescence microscope (40× objective lens) coupled to a Leica DFC340 FX camera using the Leica Application Suite v3 acquisition software.
Immunofluorescence staining of RPA and Rad51 foci
Cells were seeded onto glass chamber slides (Lab-Tek), pulsed with camptothecin (50 nmol/L) for 2 hours, and fixed at the indicated time points. Cells were immunofluorescence-stained with anti-RPA antibody (Santa Cruz) or with anti-Rad51 antibody (Calbiochem), as described previously (34, 35). Cells were subjected to in situ fractionation to clarify RPA foci, as described (36). The average numbers of RPA or Rad51 foci per nucleus were determined after scoring at least 50 nuclei.
Immunofluorescence staining of BrdUrd/ssDNA foci
Cell were grown on glass chamber slides (Lab-Tek) in the presence of BrdUrd (40 μg/mL; BD Biosciences) for one cell division cycle, pulsed with camptothecin (50 nmol/L) for 2 hours, and fixed after 8 hours. Cells were immunofluorescence-stained with anti-BrdUrd antibody (Becton Dickinson) under nondenaturing conditions (to detect BrdUrd incorporated in ssDNA), as described previously (34, 35). For clarifying BrdUrd/ssDNA foci, cells were subjected to in situ fractionation, as described (36). The average numbers of BrdUrd/ssDNA foci per nucleus were determined after scoring at least 50 nuclei.
Sister chromatid exchanges
To visualize sister chromatid exchanges (SCE), cells were grown in the presence of BrdUrd (20 μmol/L; BD Biosciences) and either DMSO or camptothecin (50 nmol/L) for 2 cell division cycles. To capture a higher number of cells in metaphase, 1 μg/mL colcemid (Sigma-Aldrich) was added 8 hours before completion of the BrdUrd treatment. Metaphase spreads were prepared, and chromosomes stained with acridine orange (Sigma-Aldrich), as described before (32). Reciprocal exchanges were counted and normalized to the total number of chromosomes per spread for 25 metaphase spreads per treatment group.
Protracted temozolomide treatment leads to acquired temozolomide resistance in glioblastoma cells in vitro
Temozolomide is part of the standard-of-care for glioblastomas; however, these tumors inevitably recur, and the recurrent tumors are resistant to further temozolomide therapy (1). Given that tumor recurrence is the main cause of mortality for patients with glioblastomas, more studies are desperately needed to address the mechanisms of temozolomide resistance in glioblastomas. As a prelude to in vivo studies, we subjected human GBM cell lines (differing in MGMT status) to protracted temozolomide exposure at a concentration of 50 μmol/L which falls within the range of the predicted concentrations of the drug in tumors in patients with glioblastomas (37). We used 2 MGMT-expressing cell lines (T98G and U138), an MGMT-negative cell line (U251), and an MGMT-negative neurosphere culture derived from a treatment-naïve patient with glioblastoma (GBM9; ref. 29). Cells were treated continuously with temozolomide in vitro for 24 days and were assessed for temozolomide sensitivity before and after the treatment period. MGMT-expressing lines (T98G and U138) were inherently resistant to temozolomide, as seen by the colony formation assay, and their temozolomide sensitivity profiles were unaltered after protracted temozolomide treatment (T98G-T12 and U138-T12; Fig. 1A and B). In contrast, the MGMT-nonexpressing lines, U251 and GBM9, were initially sensitive to temozolomide as seen by colony formation and sphere formation assays, respectively (Fig. 1C and D). Interestingly, cells surviving protracted temozolomide treatment (U251-T12 and GBM9-T12) were now significantly more resistant to the drug. However, drug resistance was not accompanied by re-expression of MGMT (7), as assessed by Western blotting of cell extracts (Fig. 1C and D), indicating that other mechanisms of temozolomide resistance may be at play. To better understand this potentially novel resistance mechanism, we decided to use an orthotopic GBM model to recapitulate acquired temozolomide resistance in vivo.
Protracted temozolomide treatment leads to acquired temozolomide resistance in glioblastoma cells in vivo
One reason for the paucity of studies addressing temozolomide resistance is the lack of matched primary and recurrent patient samples and derivative cell lines. In this study, we used an orthotopic xenograft GBM model to recapitulate and understand acquired temozolomide resistance in vivo. Of the lines we had first tested in vitro (Fig. 1), we chose GBM9 for in vivo studies because these neurosphere cultures give rise to orthotopic tumors in mice that closely recapitulate human glioblastomas in their phenotypic characteristics (29). In pilot studies, we surgically implanted GBM9 neurosphere cells into the right corpus striatum of the brains of nude mice and confirmed the presence of well-established tumors by 7 days postimplantation; there appeared to be no major variation in individual tumor growth rates, as all mice became moribund due to their tumor burden at or around 15 days postimplantation. For this study, mice bearing GBM9 tumors were treated repeatedly with temozolomide over 24 days (20 mg/kg; 12 doses given every other day), starting at 7 days postimplantation (Fig. 2A). The control group was treated with vehicle alone. Tumor-bearing mice treated with temozolomide survived significantly longer than untreated mice but eventually succumbed to their tumors, implying that the GBM9 tumors were initially responsive to temozolomide but eventually recurred (Supplementary Fig. S1A). When the mice became moribund due to their brain tumor burden, ex vivo cultures were generated from these tumors. In this model, cultures from vehicle-treated mice represent the original, treatment-naïve tumor, whereas cultures from temozolomide-treated mice represent recurrent tumors (Fig. 2A). To limit the potential problem of selection taking place during cell culturing, we used low-passage cultures (3–5 duplications) for subsequent in vitro studies. By in vitro colony survival assays, we found that the recurrent GBM9 ex vivo cultures (plotted in green) were significantly more resistant to temozolomide than the primary GBM9 ex vivo cultures (plotted in black; Fig. 2B), indicating that protracted temozolomide treatment had resulted in the generation and/or selection of drug-resistant cells. The 2 recurrent cultures that were most drug-resistant (419 and 481) were chosen for further analyses. Two treatment-naïve cultures (483 and 484) were used for comparison. The 2 sets of lines were treated with temozolomide (or with DMSO as a control), and their cell-cycle profiles were examined 24 hours after drug treatment. Temozolomide induced a robust cell-cycle block in the S–G2 phases in the primary, temozolomide-sensitive lines, as would be expected of an agent that induces replication-associated DSBs (ref. 38; Fig. 2C). In contrast, the recurrent temozolomide-resistant lines exhibited no change in cell-cycle profiles upon temozolomide treatment, indicating that they were either more proficient at repairing temozolomide-induced DSBs or had disabled G2–M checkpoints. However, when these cells were exposed to ionizing radiation (IR), we observed a robust arrest in G2 and phosphorylation of checkpoint kinases CHK1 and CHK2 (Supplementary Fig. S1B and C), demonstrating that their G2–M checkpoints were intact, and that other mechanisms are involved in temozolomide resistance.
Accelerated DNA DSB repair underlies acquired temozolomide resistance
To date, the reported mechanisms of temozolomide resistance involve either re-expression of MGMT or loss of MMR proficiency, both of which would impede the generation of DSBs in tumor cells (20). We found that neither the original GBM9 cells nor the primary or recurrent ex vivo cultures express MGMT as assessed by Western blotting (Fig. 3A). To determine MMR status, we examined 4 key MMR proteins—MSH2, MSH3, MSH6, and MLH1—whose expression levels have been reported to decrease after temozolomide treatment, resulting in drug resistance (6, 8, 39). We found no differences in protein levels between the original GMB9 cells and the sensitive or resistant ex vivo cultures, suggesting that these lines may be MMR-proficient (Fig. 3B). However, mutations in these genes have also been reported to render cells MMR-deficient (8–12). Therefore, to evaluate the MMR status of these cells, we looked for the presence of microsatellite instability (MSI), which is a functional consequence of MMR deficiency (33). MSI was assayed by amplifying 4 well-characterized unstable loci containing microsatellite repeats (33) and then determining whether the product had repeat expansions or contractions by assessing amplicon size by gel electrophoresis. We used low-passage primary human skin fibroblasts (HSF) as a wild-type reference cell line and an MMR-deficient human colon cancer line—HCT116—as a positive control (40). We found no evidence of MSI in either the original GBM9 or the ex vivo cultures derived from the tumors (note matching band patterns—black arrow—for the HSF and GBM9-derived lines and altered band pattern—red arrow—for the HCT116 line; Fig. 3C).
In agreement with these data showing lack of MGMT re-expression and proficient MMR in the recurrent cultures, we found that temozolomide induced comparable numbers of 53BP1 foci, a marker for DNA breaks, in both primary and recurrent cultures (Fig. 3D). Strikingly, the recurrent, temozolomide-resistant cultures exhibited significantly faster kinetics of resolution of 53BP1 foci in a time course experiment, indicating that augmented DSB repair might underlie the increased survival of temozolomide-resistant cells in this particular model (Fig. 3D). We reasoned that if accelerated DSB repair was the basis for acquired temozolomide resistance, then the recurrent cultures should be cross-resistant to other DSB-inducing agents, particularly agents inducing replication-associated breaks. To test this idea, we exposed cells to IR, the topoisomerase II poison etoposide, the alkylating agent MNNG, which is a functional analogue of temozolomide, or the topoisomerase I poison camptothecin. IR and etoposide induce DSBs in all phases of the cell cycle, and these breaks are primarily repaired by NHEJ (38, 41, 42). MNNG and camptothecin induce replication-associated breaks in the S-phase of the cell cycle, and these one-ended breaks can be correctly repaired only by HR (43). Interestingly, we found that both primary and recurrent cultures displayed similar sensitivities to IR or etoposide (Supplementary Fig. S2A and S2B) and repaired IR- or etoposide-induced breaks with similar kinetics (Fig. 4A and B). In contrast, the recurrent cultures were significantly more resistant than the primary cultures to MNNG or camptothecin (Supplementary Fig. S2C and S2D) and repaired MNNG- or camptothecin-induced DSBs with accelerated kinetics (Fig. 4C and D). Interestingly, the temozolomide-resistant U251-T12 cells generated in vitro (Fig. 1C) also displayed more proficient repair of temozolomide-induced DSBs compared with the parental U251 cells (Supplementary Fig. S3A) while showing no major difference in the repair of IR-induced DSBs (Supplementary Fig. S3B). Accordingly, the U251-T12 cells were more resistant to temozolomide (Fig. 1C) but were not particularly resistant to IR (Supplementary Fig. S3C). Since augmented DSB repair in the recurrent cultures was seen only with agents inducing replication-associated breaks (temozolomide, MNNG, or camptothecin), we hypothesized that the recurrent cultures might have acquired an increased capacity for HR repair which is the primary pathway utilized to repair such breaks (18).
Temozolomide-resistant cultures exhibit augmented HR repair
To examine whether augmented HR might indeed underlie temozolomide resistance in the recurrent cultures, we quantified RPA and Rad51 foci in primary versus recurrent cultures, which serve as metrics for the DNA end resection and the strand invasion steps of HR, respectively (34, 35). To accurately analyze foci formation and dissolution, we used a short 2-hour pulse of camptothecin, which is sufficient to induce DSBs in S-phase cells (as opposed to temozolomide which requires 2 rounds of DNA replication to induce breaks). Importantly, camptothecin-induced DSBs are mostly processed by HR as is evident from the fact that a high proportion of γH2AX foci induced by camptothecin colocalize with Rad51 foci in the ex vivo cultures (Supplementary Fig. S4A). The first and most critical regulatory step in HR is DNA end resection which results in the formation of 3′-OH single-stranded DNA (ssDNA) which is coated with replication protein A (RPA; ref. 44). To examine whether the temozolomide-resistant cultures were more efficient in DNA end resection, we stained the ex vivo cultures for RPA foci (an indirect measure of resection; refs. 34, 35) but found no significant difference in the numbers of camptothecin-induced RPA foci between primary and recurrent cultures (Fig. 5A). To substantiate these results, we also stained these cells for BrdUrd/ssDNA foci (a direct measure of resection; refs. 34, 35) and found no differences in the numbers of these foci (Supplementary Fig. S4B), indicating that the DNA end resection step of HR was unaltered in the recurrent cultures. We then stained for Rad51 foci to analyze the subsequent step in HR, during which Rad51 forms a nucleoprotein filament that catalyzes homology search and DNA strand exchange (45–47). While the primary and recurrent cultures displayed comparable induction of Rad51 foci, the recurrent cultures displayed faster dissolution of Rad51 foci (Fig. 5B). Dissociation of Rad51 is an essential step that exposes the 3′-end of the invading ssDNA for DNA repair synthesis and promotes downstream recombination events (45, 48). These results tentatively indicate that the mechanism underlying augmented HR in the recurrent cultures lies downstream of DNA end resection and Rad51 loading and possibly involves more efficient Rad51 dissociation leading to higher levels of recombination. Finally, we quantified the increase in recombination in the recurrent cultures by measuring SCEs occurring after treatment of cells with camptothecin (32). The recurrent cultures exhibited higher levels of SCEs compared with the primary cultures, confirming the augmented homologous recombination seen in these cultures (Fig. 5C).
Targeting HR via CDK inhibition resensitizes recurrent cultures to temozolomide
Given our results implicating augmented HR as a mechanism of resistance, we next wanted to investigate whether it might be possible to resensitize the resistant cultures to temozolomide by inhibiting HR. HR can be targeted either directly by blocking Rad51 or indirectly by subverting pathways stimulating HR (49). Although already in development, Rad51 inhibitors are not currently in clinical use (50–53). However, CDKs 1 and 2 have been shown by our group and others to regulate multiple steps during HR (35, 54–61), especially the DNA end resection step via phosphorylation of CtIP (57, 58, 60, 61) and EXO1 (35). As opposed to Rad51 inhibitors, multiple CDK 1 and 2 inhibitors are already in clinical trials (62) and could potentially be repurposed for HR inhibition. Therefore, to test whether CDK1/2 inhibitors might attenuate HR and sensitize cells to temozolomide, we used 2 CDK inhibitors—AZD5438 (a potent inhibitor of CDKs 1, 2, and 9; ref. 63) and roscovitine (a potent inhibitor of CDKs 1, 2, and 5; ref. 64). We treated the recurrent cultures with temozolomide for 48 hours (2 DNA replication cycles) to induce DSBs and then added DMSO, AZD5438, or roscovitine to the cultures. DSB repair was monitored over the next 48 hours by staining for 53BP1 foci. Unlike cells treated with DMSO, which were able to complete repair by 48 hours, cells treated with AZD5438 or roscovitine were profoundly impaired in DSB repair (Fig. 6A). Notably, this abrogation of DNA repair translated into significant sensitivity to temozolomide, as seen by the colony formation assay (Fig. 6B). Altogether, these results suggest that augmented HR can contribute to temozolomide resistance and tumor recurrence and that this can be targeted by CDK1 and 2 inhibitors for chemosensitization of recurrent tumors.
Tumor recurrence after initial treatment with radiation and temozolomide is the major cause of mortality for patients with glioblastomas, yet recurrences are understudied because of the dearth of matched tumors samples. This is of particular concern because temozolomide treatment profoundly affects the tumor's evolution as evidenced by dramatic alterations in the genomic landscape of recurrent tumors compared with their primary tumor counterparts (65). A big effort has been made recently to understand this evolutionary process by comparing the genomic and proteomic profiles of primary versus recurrent tumors and correlating these datasets with treatment regimens, radiologic data, and clinical history (65–67). However, in these studies, no functional assays were carried out to understand mechanisms of temozolomide resistance due to the lack of matching primary and recurrent cell lines. We developed an in vivo model of tumor recurrence in mice that allows us to understand temozolomide resistance by assaying matched primary and recurrent tumor cell lines. Such a model could help us identify therapy-driven changes occurring in glioma cells upon treatment with pharmacologically relevant doses of temozolomide in the context of the tumor microenvironment in an in vivo setting.
Using this model, we discovered a novel and potentially important mechanism of temozolomide resistance—augmented HR repair of temozolomide-induced DSBs. Augmented HR may be a particularly important temozolomide resistance mechanism in tumors where known resistance mechanisms (MGMT overexpression and MMR loss) do not operate. Our findings are in agreement with previous reports pointing to a role of HR in chemotherapy resistance in other cancers (49, 53, 68). Also, when we analyzed the TCGA data set (27), we found all (n = 11) recurrent glioblastomas treated with temozolomide to exhibit overexpression of HR-related genes such as Rad51. HR involves multiple steps and a large number of effector and mediator proteins (21, 22). The exact step involved in the augmented HR in our study still remains to be determined. Surprisingly, when we assessed key HR proteins (including RPA, Rad51, ATM, MRE11, NBS1, EXO1, and BRCA2) by Western blotting, we did not find any significant difference in their levels between the primary and recurrent cultures (data not shown). This suggests that the augmented HR in recurrent cultures is probably via posttranslational modifications (PTM) of HR proteins rather than by transcriptional or translational regulation. Indeed, PTMs such as phosphorylation have been shown to have essential roles in the regulation of HR enzymes like Rad51, either by modulating their enzymatic activities or intracellular localization (45). Additional mouse studies with a panel of primary glioblastoma cultures are warranted in the future to further tease apart critical HR steps and proteins that contribute to temozolomide resistance in recurrent glioblastoma.
The fact that an increase in HR is observed upon temozolomide treatment has important clinical implications. First, it suggests that rechallenging recurrent tumors with temozolomide alone may not elicit the desired therapeutic response. This is in agreement with the results of a phase II clinical trial where retreatment of recurrent tumors with temozolomide had minimal benefit for most patients, especially those who had been treated for extended periods of time with temozolomide prior to enrollment (69). Second, our results expose a druggable vulnerability—HR dependence of tumor cells resistant to temozolomide. Inhibitors of Rad51 are currently in development (50–53) and may be very effective at treating temozolomide-resistant tumors (once these drugs become clinically available). In the meantime, the novel link between CDK signaling and HR activation (70, 71) can be exploited for chemosensitization purposes, as CDK inhibitors are already in clinical trials (62) and could be quickly repurposed. Importantly, the chemosensitizing strategy tested by us (combining temozolomide with CDK inhibitors) need not be limited only to patients with recurrent glioblastomas and could work even in patients with primary glioblastomas, as long as temozolomide is able to induce DNA breaks in these tumors. This approach, besides being very toxic to all tumor cells, could be important for targeting rare temozolomide-resistant clones that may be present in the primary tumor, before they are selected for and expand under temozolomide treatment. In summary, our work uncovers a novel mechanism underlying acquired temozolomide resistance in glioblastomas and raises the possibility of improving the therapeutic response to temozolomide by targeting HR repair.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: C.R. Gil del Alcazar, B. Mukherjee, S. Burma
Development of methodology: C.R. Gil del Alcazar, S. Burma
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.R. Gil del Alcazar, P.K. Todorova, S. Burma
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.R. Gil del Alcazar, A.A. Habib, B. Mukherjee, S. Burma
Writing, review, and/or revision of the manuscript: C.R. Gil del Alcazar, P.K. Todorova, B. Mukherjee, S. Burma
Study supervision: S. Burma
S. Burma is supported by grants from the NIH (RO1CA149461, RO1CA197796 and R21CA202403) and the National Aeronautics and Space Administration (NNX16AD78G). C.R. Gil del Alcazar completed this work in partial fulfillment of the requirements for his PhD degree. A.A. Habib is supported in part by a grant from the Department of Veteran's Affairs.
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.
We thank Dr. David Boothman for valuable advice on mismatch repair assays.
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
- Received April 13, 2016.
- Revision received June 14, 2016.
- Accepted June 16, 2016.
- ©2016 American Association for Cancer Research.