Minimally Cytotoxic Doses of Temozolomide Produce Radiosensitization in Human Glioblastoma Cells Regardless of MGMT Expression

Concurrent treatment with the methylating agent temozolomide during radiotherapy has yielded the first significant improvement in the survival of adult glioblastomas (GBM) in the last three decades. However, improved survival is observed in a minority of patients, most frequently those whose tumors display CpG methylation of the O⁶-methylguanine (O⁶-meG)-DNA methyltransferase (MGMT) promoter, and adult GBMs remain invariably fatal. Some, although not all, preclinical studies have shown that temozolomide can increase radiosensitivity in GBM cells that lack MGMT, the sole activity in human cells that removes O⁶-meG from DNA. Here, we systematically examined the temozolomide dose dependence of radiation killing in established GBM cell lines that differ in ability to remove O⁶-meG or tolerate its lethality. Our results show that minimally cytotoxic doses of temozolomide can produce dose-dependent radiosensitization in MGMT-deficient cells, MGMT-proficient cells, and MGMT-deficient cells that lack mismatch repair, a process that renders cells tolerant of the lethality of O⁶-meG. In cells that either possess or lack MGMT activity, radiosensitization requires exposure to temozolomide before but not after radiation and is accompanied by formation of double-strand breaks within 45 minutes of radiation. Moreover, suppressing alkyladenine-DNA glycosylase, the only activity in human cells that excises 3-methyladenine from DNA, reduces the temozolomide dose dependence of radiosensitization, indicating that radiosensitization is mediated by 3-methyladenine as well as by O⁶-meG. These results provide novel information on which to base further mechanistic study of radiosensitization by temozolomide in human GBM cells and to develop strategies to improve the outcome of concurrent temozolomide radiotherapy. Mol Cancer Ther; 9(5); 1208–18. ©2010 AACR.

This article is featured in Highlights of This Issue, p. 1075

Adult glioblastomas (GBM; WHO grade IV astrocytomas) are invariably fatal due, in part, to resistance to postsurgical therapy. Historically, median survival following surgery and adjuvant therapy has been 9 to 12 months (1). Although the efficacy of adjuvant radiotherapy in prolonging survival has long been recognized, the benefit of alkylating agent-based chemotherapy has been equivocal (reviewed in ref. 2). Recent clinical trials have shown that median survival can be significantly prolonged and the 2-year survival rate significantly increased by administering the methylating agent temozolomide during radiotherapy and continuing temozolomide as a single agent after completion of radiotherapy (3). As a consequence, concurrent temozolomide radiotherapy is now the accepted standard of adjuvant care for newly diagnosed GBMs (4).

Temozolomide is an orally administered agent that readily crosses the blood-brain barrier and has relatively low toxicity (5). It undergoes spontaneous hydrolysis at physiologic pH to form an active metabolite that produces ∼12 base adducts in DNA (6, 7), including the cytotoxic lesions O⁶-methylguanine (O⁶-meG) and 3-methyladenine (3-meA). Improved survival following concurrent temozolomide radiotherapy is more frequent in the 40% to 45% of GBMs that exhibit promoter methylation of the gene for O⁶-meG-DNA methyltransferase (MGMT; ref. 8), indicative of epigenetic silencing of expression (9). MGMT is the sole activity that removes O⁶-meG from DNA (10), suggesting that unrepaired O⁶-meG promotes the efficacy of temozolomide radiotherapy. Yet, temozolomide radiotherapy is no more effective than radiotherapy alone in more than half of tumors that exhibit MGMT promoter methylation, whereas about 10% to 20% of GBMs lacking promoter methylation display superior outcome (3, 8), suggesting that adducts in addition to O⁶-meG may affect treatment outcome.

Although concurrent temozolomide radiotherapy has produced the first significant advance in the treatment of newly diagnosed GBMs in the last 30 years, more than half of patients derive no benefit from inclusion of temozolomide, and the prognosis remains dismal, with only a minority of patients surviving 2 years (3). Thus, there is urgent need to understand the mechanisms responsible for the efficacy of concurrent temozolomide radiotherapy to develop strategies to improve outcome and to extend the benefit of this therapy to a larger fraction of patients. To date, there is a small but growing number of preclinical studies (e.g., refs. 11–15) that address the effects of temozolomide on radiation sensitivity of human GBM cells. These studies differ widely in experimental procedures, especially in the concentration of temozolomide and the duration of temozolomide exposure before irradiation, and are notable for varying results about how temozolomide affects radiosensitivity. For example, some studies (e.g., refs. 13–15) report that temozolomide has supra-additive effects on cell killing (i.e., that temozolomide is a radiosensitizing agent), whereas others (e.g., refs. 11, 12) report simply additive effects. At present, in accord with the clinical observations noted above, enhanced radiosensitivity has been observed only in cells that express little or no MGMT. To pursue these findings, we have systematically examined the temozolomide dose dependence of radiation killing in established GBM cell lines that differ in their ability to remove O⁶-meG or to tolerate its lethality. We also examined the effect of suppressing alkyladenine-DNA glycosylase (AAG), the sole DNA repair activity in human cells that excises 3-meA (16), on killing by combined temozolomide and radiation. Our findings indicate that minimally cytotoxic doses of temozolomide can produce radiosensitization in MGMT-proficient (MGMT⁺) as well as in MGMT-deficient (MGMT⁻) GBM cell lines, and that radiosensitization is mediated by 3-meA as well as by O⁶-meG. Our results, and their possible implications, provide new insights into how temozolomide may increase radiation cytotoxicity in human GBM cells.

The human GBM-derived cell lines A1235 and A1235MR4 (hereafter MR4; provided by Dr. J. Allalunis-Turner, University of Alberta, Canada), SNB19 and SF767 (Brain Tumor Research Laboratory, University of California, San Francisco), and T98G (American Type Culture Collection) were grown as previously described (17).

We have previously described in detail the procedures for use of temozolomide and O⁶-benzylguanine (O⁶-BG), and the determination of proliferative survival by clonogenic assay (17–19). Briefly, unless otherwise stated, 1,000 to 2,000 cells were incubated for 2 hours with temozolomide before ¹³⁷Cs-γ-ray irradiation at 1 Gy/min under ambient conditions. Incubation was continued for 22 hours after irradiation before changing to fresh medium to allow the formation of colonies ≥50 cells. Survival (mean ± SD) is the ratio of colony-forming ability of treated cells to that of nontreated cells. For experiments defining the temozolomide dose dependence of enhanced radiation cytotoxicity, radiosensitization is the ratio of the survival expected if the effects of temozolomide and radiation were additive to the survival observed. For example, if treatment with temozolomide alone and γ-rays alone produces 95% and 80% survival, respectively, whereas treatment with both temozolomide and γ-rays reduces survival to 45%, radiosensitization would be 75% divided by 45% or 1.7-fold. For full γ-ray survival curves, cytotoxicity was quantitated by linear regression analysis of plots of log surviving fraction versus radiation dose to obtain the three resistance parameters, LD₁₀, D_T, and D₃₇, as we have previously described in detail (18). Survival was determined in three separate experiments in which every dose was assayed in triplicate (i.e., nine determinations per dose) to achieve statistical significance.

γ-H2AX content of 50,000 to 200,000 cells solubilized in Laemmli buffer was estimated by Western blotting (20). Detection was by chemiluminescence using standard techniques; a digital camera imaging system was used to produce digital images of blots for analysis of signal intensity. γ-H2AX signal intensity was normalized to that of β-actin, as a loading control. The ratio was then normalized to that for nontreated cells, a control for γ-H2AX expression due to endogenous processes (e.g., mitosis; ref. 21), to permit comparison between separate determinations. Values are the mean ± SD of four experiments.

We have previously provided detailed description of antisense (ASO) and sense oligonucleotides (SO) targeting AAG, cationic lipid–mediated transfection, and controls for specificity (17). Forty-eight hours after transfection, when suppression of activity is maximal (17), cells were subcultured for cytotoxicity analyses.

Data analysis and statistical procedures were done using Microsoft Excel. Comparison of means was by Student's t test assuming unequal variances. Relationships between continuous variables were assessed by regression analysis. Statistically significant relationships were determined at the 95% confidence level.

We began this study by examining the temozolomide dose dependence of sensitization to γ-rays of two human GBM lines that express no detectable MGMT activity (<0.25 fmol/1 × 10⁶ cells; refs. 17, 18). These lines, SNB19 and A1235, are highly sensitive to temozolomide, with LD₁₀s for a 24-hour exposure of 37 ± 4 μmol/L and 23 ± 2 μmol/L, respectively.³

In these experiments, we sought to mimic clinical practice in which a single daily treatment consists of oral temozolomide at 75 mg/m² taken shortly before treatment with 2 Gy radiation (4). We therefore incubated cells with temozolomide for 2 hours, the interval reported to produce peak cerebral spinal fluid concentration in nonhuman primates (22), before radiation with 2 Gy ¹³⁷Cs γ-rays. As the half-life of temozolomide in aqueous solution is similar to that in serum in vivo (1.5 versus 1.8 h; ref. 23), we continued incubation in the presence of temozolomide for 22 hours after irradiation to simulate exposure during a single-treatment fraction. To further approximate conditions that may prevail in situ, we used temozolomide doses within the range attainable in cerebral spinal fluid, i.e., 10 to 25 μmol/L (22, 24).

Figure 1A shows that incubating SNB19 cells for 2 hours with temozolomide doses up to 10 μmol/L had little effect on cell survival (>90%), whereas 2 Gy γ-rays alone reduced survival to 81 ± 4%. Figure 1A and B show that treatment with minimally cytotoxic doses of temozolomide in combination with radiation produced supra-additive killing that was temozolomide dose dependent. At 10 μmol/L temozolomide, survival of combined treatment was decreased 1.8- ± 0.2-fold relative to radiation alone (81 ± 4% versus 46 ± 9%; P ≤ 0.001; Table 1). However, at 15 μmol/L temozolomide, a dose that reduced survival to ∼40% (Fig. 1A), the enhancement of radiation killing was diminished to 1.2-fold (Fig. 1B). A similar pattern of temozolomide-mediated sensitization to killing by 2 Gy γ-rays was observed for A1235 cells (Fig. 1C and D), with maximal enhancement observed at 5 μmol/L temozolomide (Table 1). These data show that nonlethal or minimally cytotoxic doses of temozolomide can sensitize established MGMT⁻ cell lines to killing by 2 Gy γ-rays. The observations are important because the temozolomide concentrations that produced supra-additive killing are likely attainable in GBM cells in situ (24).

Mismatch repair (MMR) mediates the cytotoxicity of O⁶-meG and inactivation of MMR renders cells insensitive to killing by this adduct (10). To investigate the possibility that MMR contributes to temozolomide-mediated radiosensitization, we examined MR4 cells. MR4 is a well-characterized human GBM cell line that lacks both MGMT and MMR activities (25). It was derived from MGMT⁻ A1235 cells by selection for methylation resistance (26); in accord, the LD₁₀ for a 24-hour exposure to temozolomide, 1,339 ± 77 μmol/L, is 58-fold greater than that for the parental A1235 line.³ As shown in Fig. 1E, MR4 cells are insensitive to temozolomide at doses as high as 200 μmol/L, whereas exposure to 2 Gy γ-rays reduced survival to 77 ± 8%. Treatment with temozolomide at doses that sensitized A1235 and SNB19 cells to radiation (i.e., ≤ 10 μmol/L) had no discernible effect on γ-ray killing in MR4; however, increased sensitivity was detectable at temozolomide concentrations of ≥25 μmol/L (Fig. 1E and F). Treatment with 100 μmol/L temozolomide and 2 Gy γ-rays reduced survival 1.8- ± 0.3-fold compared with radiation alone (77 ± 8% versus 43 ± 7%; P ≤ 2 × 10⁻⁶; Table 1). This finding indicates that MMR mediates radiosensitization by temozolomide in MGMT⁻ GBM cells.

Based on the temozolomide dose dependence of radiosensitization of MGMT⁻MMR⁻ MR4 cells (Fig. 1E and F), we examined the effect of a range of minimally cytotoxic temozolomide doses on radiation killing in the MGMT⁺ GBM line SF767. SF767 cells contain 61 ± 12 fmol/1 × 10⁶ cells (i.e., ∼37,000 molecules/cell) of MGMT (18) and exhibit an LD₁₀ of 1,119 ± 62 μmol/L for a 24-hour temozolomide exposure.³ Figure 1G illustrates the effect of temozolomide dose on killing following irradiation with 2 Gy  γ-rays. Temozolomide alone at doses up to 100 μmol/L did not detectably reduce survival, reflecting the temozolomide resistance of MGMT⁺ cells, whereas radiation alone reduced survival to 76% ± 6%. Treatment with both temozolomide and γ-rays produced supra-additive killing that was temozolomide dose dependent (Fig. 1H). Relative to radiation alone, survival was decreased 1.7- ± 0.2-fold at 100 μmol/L (76 ± 6% versus 45 ± 5%; P ≤ 2 × 10⁻⁶; Table 1). Temozolomide concentrations that sensitized MGMT⁻ cells to radiation (e.g., ≤10 μmol/L; Fig. 1A–D) had little or no effect on γ-ray killing. We also observed temozolomide-mediated radiosensitization in the MGMT⁺ GBM line T98G (Table 1; Fig. 1K and L), with maximal enhancement of killing between 50 and 100 μmol/L temozolomide. The results for T98G are notable in that this line is insensitive to killing by 2 Gy. These data show that essentially nonlethal doses of temozolomide can produce supra-additive radiation cytotoxicity in MGMT⁺ GBM cells, albeit at concentrations higher than required in MGMT⁻ cells.

To quantify the contribution of MGMT to protection of SF767 cells from temozolomide-enhanced radiation killing, we ablated MGMT activity with the substrate analogue inhibitor O⁶-BG. Treatment with O⁶-BG reduces MGMT to undetectable levels (<0.25 fmol/1 × 10⁶ cells) and decreases LD₁₀ for temozolomide about 12-fold (1,119 ± 62 versus 91 ± 6 μmol/L).³ As shown in Fig. 1I, ablating MGMT activity greatly reduced the temozolomide dose required to produce radiosensitization (compare the X-axes in Fig. 1G versus I). Exposure to temozolomide alone had only a small effect on survival (>90% at all doses), whereas exposure to radiation alone reduced survival to 77 ± 7%. Relative to cells receiving radiation only, maximal decrease in survival, 1.7- ± 0.2-fold, was observed at doses as low as 5 μmol/L temozolomide (77 ± 7% versus 45 ± 7%; P ≤ 0.0004; Table 1; Fig. 1J). These results are similar to those for the MGMT⁻ lines SNB19 and A1235 (Fig. 1A–D), and strongly indicate that unrepaired O⁶-meG plays a prominent role in enhancing radiation-induced cytotoxicity.

To further characterize the effect of temozolomide on radiation killing, we examined the dose dependence of γ-ray survival of MGMT⁻ SNB19, MGMT⁺ SF767, and MGMT⁻MMR⁻ MR4 cells in the absence and presence of temozolomide (Fig. 2). In the absence of temozolomide, the lines differ little in radiosensitivity, displaying similar values for LD₁₀, the dose that reduces survival to 10%, for D_T, the threshold dose that defines a shoulder of resistance below which cells are insensitive to killing, and for D₃₇, a measure of the rate of cell killing. In these three lines, exposure to minimally cytotoxic concentrations of temozolomide (10 μmol/L for SNB19 and 100 μmol/L for SF767 and MR4; Table 1) for 2 hours before and 22 hours after irradiation increased radiosensitivity as evidenced by statistically significant 1.5- to 1.9-fold reductions in LD₁₀ (Fig. 2). Lower LD₁₀ reflected (a) a large 7.2- to 17-fold reduction in D_T and (b) a 1.5- to 2.0-fold reduction in D₃₇. These data provide further strong evidence that temozolomide can act as a radiosensitizing agent and, together with the data for MGMT⁺ (Fig. 1G, H, K, and L) and MGMT⁻MMR⁻ cells in Fig. 1E and F, show that removal or tolerance of O⁶-meG does not preclude radiosensitization.

It has been reported that temozolomide-enhanced killing of MGMT⁻ GBM cells requires exposure to the drug before radiation (e.g., refs. 13–15). To determine whether this is the case for MGMT⁺ cells, we analyzed SF767 cells treated with 100 μmol/L temozolomide for different intervals before and after irradiation. As shown in Table 2, temozolomide alone did not reduce survival and 2 Gy γ-rays alone reduced survival to 76 ± 0.3%. Incubation with 100 μmol/L temozolomide for 2 hours before and for 22 hours after radiation reduced survival to 46 ± 3%, relative to radiation alone, a statistically significant reduction (P ≤ 0.005). Limiting temozolomide treatment to 2 hours before radiation reduced survival to the same extent (48 ± 7%), whereas exposure to temozolomide for 22 hours immediately after radiation did not reduce survival relative to radiation alone (75 ± 3%). As also shown in Table 2, essentially the same results were observed for SF767 cells treated with 5 μmol/L temozolomide and O⁶-BG. These data provide evidence that radiosensitization requires the action of temozolomide before irradiation in MGMT⁺ cells as well as cells that lack MGMT and that temozolomide exposure after irradiation does not produce supra-additive killing. They also suggest that methylation damage that would otherwise be repaired or tolerated is converted to lethal lesions by γ-rays.

Histone H2AX is rapidly phosphorylated at serine 139 to form γ-H2AX in response to double-strand breaks (DSB) produced by ionizing radiation (21). γ-H2AX functions in the cellular response to DSBs by promoting chromatin remodeling and assembly of repair proteins, and serves as a sensitive marker of DSBs. Figure 3A shows a representative determination of γ-H2AX and β-actin content, assessed by Western blotting of extracts of MGMT⁺ SF767 following no treatment; incubation for 2 hours with 100 μmol/L temozolomide; irradiation with 2 Gy γ-rays alone; or incubation for 2 hours with 100 μmol/L temozolomide before 2 Gy γ-rays. For all treatments, cells were harvested and extracts were prepared about 30 to 45 minutes after irradiation. Digital image analysis of this blot together with that of three additional experiments (Fig. 3B) revealed that γ-H2AX content, normalized to actin, was the same in cells treated with temozolomide alone as in nontreated cells. In cells exposed to 2 Gy γ-rays alone, γ-H2AX content was 2.0- ± 0.6-fold greater than in nontreated cells, whereas the content of cells treated with temozolomide and γ-rays was 2.8- ± 0.8-fold greater, a difference that is statistically significant (P < 0.02). Comparable results were also observed for SF767 treated with O⁶-BG (Fig. 3C), the MGMT⁺ line T98G (Fig. 3D), and the MGMT⁻MMR⁻ line MR4 (Fig. 3E). In all cases, the difference in γ-H2AX content between cells treated with radiation versus temozolomide plus radiation was statistically significant (P ≤ 0.05). These findings indicate that rapid elevation of DSB content accompanies enhanced sensitivity to radiation killing in temozolomide-treated GBM cells.

Our observation that temozolomide increases γ-ray sensitivity in MGMT⁺ and MGMT⁻MMR⁻ GBM cells at doses 10- to 20-fold greater than those required in MGMT⁻ cells suggests that a temozolomide-induced adduct in addition to O⁶-meG may contribute to radiosensitization. To assess this possibility, we determined the effect of suppressing the DNA repair enzyme AAG on the temozolomide dose dependence of radiosensitization. The preferred substrate of AAG is the cytotoxic lesion 3-meA (16), which comprises ∼10% of temozolomide base adducts (6, 7). AAG is the only repair activity in human cells that is known to excise 3-meA from DNA (16). We have previously reported (17) that suppressing AAG activity increases the sensitivity of human GBM cells, including SF767 and MR4, to methyl-lexitropsin, an agent that produces almost exclusively 3-meA in DNA (27). The dose dependence of temozolomide-induced sensitization to 2 Gy γ-rays in MGMT⁺ SF767 transfected with ASO targeting AAG or SO is shown in Fig. 4. Transfection with SO had little or no effect on either killing by temozolomide or γ-rays alone, or on temozolomide-enhanced γ-ray killing compared with untransfected cells (compare Fig. 4A and C versus Fig. 1G and H). Transfection with ASO, which reduces AAG activity 1.8-fold (17), had no effect on radiation killing and only a modest effect on temozolomide sensitivity, reducing survival at 100 μmol/L from 100% to 90% (compare Fig. 4A versus B). ASO treatment, however, did produce 1.2- to 1.4-fold, statistically significant (P ≤ 0.05) decreases in survival (Fig. 4A versus B) compared with SO-treated cells at temozolomide doses ≥25 μmol/L. The effect of ASO is more clearly illustrated by the statistically significant (P ≤ 2 × 10⁻⁶; ANOVA) 1.5- to 2-fold reductions in temozolomide dose dependence for radiosensitization that accompanied ASO treatment (Fig. 4C): for example, the temozolomide dose required to produce a 1.5-fold sensitization was 1.8-fold lower (55 versus 100 μmol/L) in ASO- versus SO-treated cells. Comparable results were observed for MGMT⁻MMR⁻ MR4 cells (Fig. 4D–F), in which the temozolomide dose required to produce a 1.5-fold radiosensitization was 2.5-fold lower (30 versus 75 μmol/L) in ASO- versus SO-treated cells. These findings indicate that reduced excision of 3-meA can contribute to temozolomide-induced radiosensitization in human GBM cells.

The long-term goal of the present work is to better understand the mechanisms underlying the improved efficacy of concurrent temozolomide radiotherapy, relative to radiotherapy alone, in the adjuvant care of GBMs. Toward this end, we compared the temozolomide dose dependence of sensitization to killing by γ-rays in GBM cell lines that differ in their ability to remove or tolerate O⁶-meG. Our results show that (a) minimally cytotoxic doses of temozolomide can produce supra-additive cytotoxicity, suggestive of an interaction between methylation- and radiation-induced DNA damage; (b) temozolomide can enhance radiosensitivity in MGMT⁻, MGMT⁻MMR⁻, and MGMT⁺ GBM cells, the dose dependence being 10- to 20-fold lower in MGMT⁻ cells, indicative of a role for O⁶-meG and MMR in radiosensitization; (c) enhanced radiation killing requires treatment with temozolomide before, but not after, irradiation and is accompanied by elevated DSB content within 45 minutes after irradiation; (d) suppressing AAG increases radiosensitivity, indicating that the DNA adduct 3-meA can promote radiation cytotoxicity. These findings strongly support the conclusion that temozolomide is a radiosensitizing agent and suggest that a common mechanism(s) of radiosensitization may operate in MGMT⁺, MGMT⁻MMR⁻, and MGMT⁻ GBM cell lines.

In accord with previous studies (e.g., refs. 13–15), we observed radiosensitization by temozolomide in two GBM lines that do not express MGMT (Table 1; Fig. 1A–D). In the absence of MGMT activity, maximal enhancement of killing by 2 Gy γ-rays, the standard fractionated radiotherapy dose, was observed at temozolomide doses below those attainable in cerebral spinal fluid (i.e., 10–25 μmol/L; refs. 22, 24). In fact, an increase in radiation killing was detectable with as little as 2.5 μmol/L temozolomide, suggesting that unrepaired O⁶-meG is a potent sensitizing lesion. This conclusion is supported by the finding that ablating MGMT activity in MGMT⁺ cells with O⁶-BG reduces the temozolomide dose dependence for radiosensitization by 20-fold (Fig. 1G and H versus I and J). Additional evidence for a role for O⁶-meG is provided by the 20-fold increase in the temozolomide dose required for radiosensitization in the MGMT⁻MMR⁻ line MR4, relative to its MGMT⁻ A1235 parent line (Fig. 1C and D versus E and F). This latter observation indicates that radiosensitization by O⁶-meG, like the cytotoxicity of this lesion (10), involves the activity of MMR. Lastly, we observed radiosensitization in two MGMT⁺ lines (Fig. 1G, H, K, and L), at temozolomide doses at least 10 times greater than required in MGMT⁻ cells (Table 1). The higher temozolomide doses may be necessary to produce enough O⁶-meG and/or additional radiosensitizing lesions to escape repair. The foregoing results provide mechanistic support for the clinical observation that GBMs displaying CpG methylation in the MGMT promoter, a presumptive marker of gene silencing (9), are more likely to benefit from concurrent temozolomide radiotherapy (3, 8).

With one exception (13), our studies differ from previous work in that we examined the interaction of minimally cytotoxic doses of temozolomide (i.e., <10% reduction in survival) with γ-radiation, an experimental approach that greatly facilitated the detection of radiosensitization. Our data show a decrease in the enhancement of radiation killing when the contribution of temozolomide to cytotoxicity exceeds 20% (Fig. 1A–D, K, and L). These findings suggest that in our experimental protocol, higher doses of temozolomide may obscure the detection of radiosensitization by producing a high background of cell killing by temozolomide alone; some earlier studies that did not observe radiosensitization in some MGMT⁻ GBM lines (e.g., refs. 11, 14) used temozolomide concentrations that reduced survival by 40% or greater.

Our data indicate that 3-meA is a radiosensitizing adduct and that temozolomide-induced potentiation of radiation killing can occur independently of O⁶-meG. These conclusions are based on the statistically significant reduction in the temozolomide dose required for radiosensitization that accompanies the transfection of MGMT⁺ SF767 and MGMT⁻MMR⁻ MR4 cells with antisense directed against AAG (Fig. 4). Although AAG is the only DNA repair activity in human cells that excises 3-meA, it has a broad substrate specificity, excising a variety of alkyl and oxidative base lesions, although less efficiently than 3-meA in almost all cases (16). Thus, it is possible that increased radiosensitization in antisense-treated cells may reflect, in part, the action of additional lesions.

The tumoricidal activity of radiation and clinically used alkylating agents is mediated almost exclusively by DSBs (10, 28, 29). Hence, the goal of therapy with these and other DNA-damaging agents is to cause sufficient DSBs to exceed the repair capacity of tumor cells. Current evidence indicates that the cytotoxicity of O⁶-meG and 3-meA is mediated by the interaction of these adducts with DNA replication (10), suggesting a parsimonious explanation of how exposure to temozolomide before irradiation may yield supra-additive cytotoxicity in GBM cell lines: O⁶-meG and 3-meA produce single-strand gaps that are converted into DSBs upon subsequent irradiation. In the case of O⁶-meG, recognition of O⁶-meG:T mispairs by MMR is followed by exonucleolytic degradation, extending from the 3′ terminus of the newly replicated strand to a position 5′ to the mispair (30). It has been proposed that repair synthesis to fill the gap produces another mispair, thus eliciting another round of excision (reviewed in ref. 10). This “futile cycle” would in effect produce a persistent single-strand gap. In the case of 3-meA, diverse evidence strongly indicates that this adduct impedes DNA replication fork progression (31–34), which can result in single-strand gaps, as shown in yeast and frog oocytes (35, 36), and in mammalian cells (37). Additional evidence that O⁶-meG and 3-meA are the precursors of single-strand gaps is that both adducts elicit sister chromatid exchange, a hallmark of DNA gap filling (10). The single-strand gaps produced by the action of MMR or blocked replication fork progression are believed to range in length from ∼150 nucleotides to >300 nucleotides, respectively (30, 36). Gaps of this size would greatly increase the vulnerability of DNA to DSB formation because strand scission by radiation-induced oxidative free radicals at any nucleotide in the single-stranded region converts the gap into a DSB. The resulting susceptibility to DSB formation could account for the supra-additive increases in γ-ray cytotoxicity that accompany treatment with minimally cytotoxic doses of temozolomide.

The foregoing scenario provides a simple mechanism to account for the elevation of DSB content observed shortly after irradiation (Fig. 3) and the requirement that temozolomide exposure precede irradiation. It also offers a facile explanation for the greater temozolomide dose dependence of radiosensitization observed in MGMT⁻MMR⁻ MR4 compared with its MGMT⁻ parental line A1235, i.e., the inability of O⁶-meG to produce gaps in the absence of MMR necessitates greater concentrations of temozolomide to yield compensatory levels of 3-meA and other radiosensitizing lesions. Conversely, the reduction of temozolomide dose dependence accompanying the ablation of MGMT or transfection with anti-AAG antisense reflects the greater likelihood of a DNA replication fork encountering O⁶-meG or 3-meA in repair-compromised cells. Finally, the scenario accounts for the large, temozolomide-induced reduction in the radiation dose tolerated without cytotoxicity, exhibited as the shoulder on the survival curves in Fig. 2 and quantitated as the parameter D_T, by providing a mechanism by which otherwise innocuous levels of methylation-induced single-strand gaps are converted into levels of DSBs that saturate repair capacity.

Our findings do not rule out the possibility that the radiosensitizing effects of temozolomide reflect decreased repair of radiation-induced DSBs or other radiation lesions such as single-strand breaks as has been suggested in earlier studies (13, 14). Conceivably, the presence of O⁶-meG or 3-meA near the end of DSBs may prevent damage recognition or processing. Such a mechanism is consistent with the large decrease in D_T mentioned above. However, it is difficult to imagine that the minimally cytotoxic doses of temozolomide would yield enough adducts to interact with the small number of DSBs produced by 2 Gy γ-rays (approximately 75–150 per diploid nucleus; ref. 38). Moreover, decreased repair does not necessitate exposure to temozolomide before irradiation and does not explain the large increase in temozolomide dose dependence that accompanies the loss of MMR in MGMT⁻ cells.

Our studies suggest that resistance to temozolomide radiotherapy is multifactorial and that MMR and AAG, in addition to MGMT, may influence treatment outcome. Repeated episodes of methylation strongly select for MMR-deficient variants of GBM cell lines that do not express MGMT (e.g., ref. 26) and accumulating evidence suggests that recurrence in GBMs treated with temozolomide radiotherapy can be accompanied by epigenetic and mutational inactivation of MMR (e.g., refs. 39, 40). Notably, MMR-defective variants arising in MGMT⁻ GBM cells before treatment with temozolomide radiotherapy would likely display CpG methylation of the MGMT promoter and our findings with MR4 suggest that promoter methylation in the absence of MMR might lead to false prediction of favorable clinical response.

The mechanistic scenario described above suggests several additional potential mechanisms of resistance to temozolomide radiotherapy. These include the activation of the intra–S-phase checkpoint (41) in response to the single-strand gaps generated by O⁶-meG and 3-meA. Checkpoint activation reduces the rate of firing of new origins of replication and slows the rate of elongation of ongoing forks. Intra–S-phase arrest would promote resistance to temozolomide radiotherapy by providing additional time for adduct removal from template DNA, thereby reducing encounters of DNA replication with methyl adducts. Processes that mediate gap filling and replication restart at stalled forks are other potential resistance mechanisms (35, 42). Among the latter, the Werner syndrome helicase promotes the recovery of stalled replication forks (42) and increases resistance to O⁶-meG and 3-meA in human GBM cells (19).

Circumventing DNA repair–mediated resistance to temozolomide is likely to increase the clinical efficacy of concurrent temozolomide radiotherapy and our findings suggest that targeting the repair of O⁶-meG and/or 3-meA may be promising antiresistance strategies. Inhibiting the removal of O⁶-meG is an important priority. Clinical trials have shown that O⁶-BG can deplete MGMT activity in high-grade gliomas and additional trials are now ongoing to evaluate the effect on the outcome of ablating MGMT during alkylating agent treatment (reviewed in ref. 4). Our data suggest that inhibiting the repair of 3-meA may complement the depletion of MGMT in overcoming resistance to temozolomide. Repair of 3-meA is an especially attractive antiresistance target in MGMT⁻ GBM cells that have acquired temozolomide resistance through the inactivation of MMR. Base excision repair pathways (10) that process 3-meA present multiple targets for antiresistance intervention. The development of inhibitors of AAG and other members of base excision repair pathways, e.g., Ape1/Ref-1 and poly(ADP-ribose) polymerase, is ongoing (4, 43). Inhibition of poly(ADP-ribose) polymerase has been shown to sensitize human GBM xenografts to temozolomide alone (44) and to temozolomide used with radiation (45).

No potential conflicts of interest were disclosed.

We dedicate this article with respect and affection to the memory of our colleague and friend Dr. Alexander M. Spence.

Grant support: American Cancer Society grant RSG 01 9101 CCE (M.S. Bobola) and NIH grant number CA82622 and CA104593 (J.R. Silber).

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.