O⁶-Methylguanine-DNA Methyltransferase, O⁶-Benzylguanine, and Resistance to Clinical Alkylators in Pediatric Primary Brain Tumor Cell Lines

Purpose: Primary brain tumors are the leading cause of cancer death in children. Our purpose is (a) to assess the contribution of the DNA repair protein O⁶-methylguanine-DNA methyltransferase (MGMT) to the resistance of pediatric brain tumor cell lines to clinical alkylating agents and (b) to evaluate variables for maximal potentiation of cell killing by the MGMT inhibitor O⁶-benzylguanine, currently in clinical trials. Few such data for pediatric glioma lines, particularly those from low-grade tumors, are currently available.

Experimental design: We used clonogenic assays of proliferative survival to quantitate cytoxicity of the chloroethylating agent 1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU) and the methylating agent temozolomide in 11 glioma and five medulloblastoma lines. Twelve lines are newly established and characterized here, nine of them from low-grade gliomas including pilocytic astrocytomas.

Results: (a) MGMT is a major determinant of BCNU resistance and the predominant determinant of temozolomide resistance in both our glioma and medulloblastoma lines. On average, O⁶-benzylguanine reduced LD₁₀ for BCNU and temozolomide, 2.6- and 26-fold, respectively, in 15 MGMT-expressing lines. (b) O⁶-Benzylguanine reduced D_T (the threshold dose for killing) for BCNU and temozolomide, 3.3- and 138-fold, respectively. D_T was decreased from levels higher than, to levels below, clinically achievable plasma doses for both alkylators. (c) Maximal potentiation by O⁶-benzylguanine required complete and prolonged suppression of MGMT.

Conclusions: Our results support the use of O⁶-benzylguanine to achieve full benefit of alkylating agents, particularly temozolomide, in the chemotherapy of pediatric brain tumors.

Primary brain tumors are the most common solid malignancy of childhood, with 2,200 new cases diagnosed annually (1). The majority (60-70%) are gliomas (astrocytomas, oligodendrogliomas, and ependymomas) histologically similar to those found in adults (2). The remainder consists of diagnostic types uncommon in adults, including medulloblastoma, primitive neuroectodermal tumors, and mixed neuronal-glial tumors (1, 2). Whereas adult tumors occur predominantly in the cerebral hemispheres, half of pediatric cases occur in the cerebellum and brain stem (2). The contemporary standard of care for malignant and many subtotally resected low-grade pediatric brain tumors includes post-operative radiation therapy and/or multiagent chemotherapy. This treatment strategy has produced dramatic increases in 5-year survival rates for the majority of medulloblastomas (3) but has been less effective in improving the prognosis for malignant gliomas and for tumors in infants and young children. In addition, there are no effective therapies for most tumors that recur after previous radiation and chemotherapy. Thus, primary brain tumors are the leading cause of cancer death in children, the overall 5-year survival rate being 50% (1).

Chloroethylating agents [e.g., 1-(2-chloroethyl)-3-cyclohexyl-1-nitrosourea (CCNU), 1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU)] and methylating agents (e.g., temozolomide, procarbazine), used in single-agent or combination regimens, are key components in the chemotherapy of pediatric brain tumors (e.g., refs. 4, 5). Among the most effective drugs for treatment of primary brain tumors, these agents produce a diversity of alkyl adducts in DNA (6–8). Their cytotoxicity has been definitively associated with alkylation at the O⁶ atom of guanine (6, 9). Chloroethylating agents introduce O⁶-chloroethylguanine, a precursor of the lethal interstrand cross-link 1-(3-cytosinyl), 2-(1-guanyl)ethane, whereas methylating agents introduce the monoadduct O⁶-methylguanine. Current evidence indicates that persistent interstrand cross-links and O⁶-methylguanine impede DNA replication, resulting in lethal double-strand breaks at collapsed replication forks (10).

A large body of work with brain tumor-derived cell lines and xenografts has shown that the DNA repair protein O⁶-methylguanine-DNA methyltransferase (MGMT) contributes to alkylating agent resistance (e.g., refs. 11–16). MGMT catalyzes the transfer of simple, branched, and halogenated alkyl groups from the O⁶ position of guanine in double-stranded DNA to an internal cysteine, yielding guanine and S-alkylcysteine (17). Because the alkyl receptor site is not regenerated, the number of O⁶-alkylguanine adducts that can be removed from DNA in vivo is limited by the number of MGMT molecules and the rate of synthesis of the protein. The majority of adult (18, 19) and pediatric (20) primary brain tumors express MGMT activity. Activity in tumors is elevated 2- to >500-fold relative to adjacent normal brain in ∼65% of adult and pediatric cases (18–20), raising the possibility that MGMT contributes to tumor alkylator resistance in vivo. In accord, low MGMT content, assessed by immunohistochemistry or inferred from the methylation status of the MGMT promoter, has been associated with better clinical outcome following alkylating agent-based chemotherapy in adult gliomas (reviewed in ref. 21).

Ablation of MGMT activity with the substrate analogue inhibitor O⁶-benzylguanine (21, 22) enhances the cytotoxicity of methylating and chloroethylating agents in primary brain tumor cell lines (e.g., refs. 11–14) and xenografts (e.g., refs. 15, 16). These preclinical studies are central to the development of treatment regimens to evaluate the efficacy O⁶-benzylguanine in improving response to alkylator-based chemotherapy in adult gliomas (21, 22) and in pediatric brain tumors (e.g., PBTC-005; http://www.cancer.gov). Extensive studies with adult glioma lines have revealed both heterogeneity in benefit conferred by O⁶-benzylguanine, and a requirement for prolonged incubation with O⁶-benzylguanine after alkylator exposure to realize maximal potentiation of cell killing. Comparable evaluation of the efficacy of O⁶-benzylguanine in suppressing alkylator resistance in pediatric brain tumor cells has been less thorough due partly to a lack of cell lines, especially those derived from gliomas. Here we examine the contribution of MGMT to BCNU and temozolomide resistance in 16 pediatric brain tumor-derived cell lines, including 12 lines that we have newly characterized. Nine of the new lines were derived from low-grade gliomas for which there is a paucity of data on alkylator sensitivity and response to O⁶-benzylguanine. Our results show that MGMT is a major determinant of BCNU resistance and the predominant determinant of temozolomide resistance in cultured pediatric glioma and medulloblastoma cells. The data also show that maximal suppression of resistance to both agents requires ablation of MGMT activity with O⁶-benzylguanine not only before but also for a prolonged period after alkylator exposure. Our results support the use of O⁶-benzylguanine to achieve full benefit of alkylating agent chemotherapy for pediatric brain tumors.

Establishment and characterization of cell lines. Tumors and demographic information were obtained from informed patients according to protocols approved by the Institutional Review Board at Children's Hospital and Regional Medical Center. Diagnosis was obtained from the final neuropathology report. The 12 new cell lines were established as previously described with minor modifications (23). Briefly, tumors were transported from the operating room in ice-cold DMEM/F12 containing 5% iron-supplemented bovine serum, 100 units/mL penicillin, 100 μg/mL streptomycin, and 0.25 μg/mL amphotericin B. After removing large blood vessels and necrotic material, specimens were washed repeatedly with sterile, ice-cold PBS. The tissue was minced with scalpel blades in supplemented medium and serially passed through 18-, 20-, and 22-gauge needles to produce a single cell suspension, as verified by microscopic examination. The cells were pelleted by centrifugation at 800 × g and resuspended in 10 mL of 17 mmol/L Tris-HCl (pH 7.2), 140 mmol/L NH₄Cl. After incubation at 37°C for 10 minutes to lyse erythrocytes, the cells were washed with PBS. The washed cell pellet was resuspended in PBS and viable cell titer was determined by trypan blue exclusion using a hemacytometer. Supplemented medium was inoculated with ≥ 2 × 10⁶ viable cells, and cultures were incubated at 37°C in 5% CO₂/95% humidified air. Proliferation of tumor cells was evident within 10 to 21 days as foci of overgrown cells. The establishment, phenotypic characterization, and contribution of MGMT to alkylator resistance have been previously reported for the medulloblastoma lines UW228-1, UW228-2, and UW228-3 and the glioma line UW467 (11–13, 23).

All lines proliferate as adherent monolayers that have been maintained in continuous culture for >50 to 200 passages and readily form colonies (20-30% plating efficiency). The lines do not display contact inhibition at high cell density and grow as spheroids in 0.25% agar and in suspension culture. As shown in Table 1, immunohistochemistry revealed expression of glial fibrillary acidic protein, S100 protein, synaptophysin, neuron-specific nuclear protein, and/or neuron-specific enolase, antigenic markers frequently expressed by tumors and cell lines of neuroepithelial origin (refs. 24, 25 and references therein) in all but one line.

Treatment with O⁶-benzylguanine. Cells were incubated with O⁶-benzylguanine (a gift from Dr. Robert Moschel, Laboratory of Comparative Carcinogenesis, National Cancer Institute, Frederick, MD) to eliminate MGMT activity. Unless otherwise indicated, cells were plated in the presence of 20 μmol/L O⁶-benzylguanine or an equivalent volume of its solvent DMSO and incubated for 18 to 20 hours to allow attachment and resumption of proliferation. The cells were maintained in 20 μmol/L O⁶-benzylguanine during the subsequent exposure to cytotoxic drugs (see above) and for 18 to 20 hours after, as previously described (11). Final DMSO concentration added with O⁶-benzylguanine was <0.1%.

Drug sensitivity. BCNU and temozolomide were obtained from the pharmacies of the University of Washington Medical Center and Children's Hospital and Regional Medical Center, respectively. BCNU was dissolved in absolute ethanol whereas temozolomide, and N-methyl-N′-nitro-N-nitrosoguanidine (MNNG, Sigma, St. Louis, MO) were dissolved in DMSO. All drugs were stored as single-use aliquots at −80°C. Immediately before use, stock solutions were diluted in solvent so that a constant volume was added for all drug doses. The final concentration of ethanol added with BCNU was <0.5% and of DMSO added with temozolomide or MNNG was 0.5%. Controls received solvent only.

Drug sensitivity was quantitated by assay of clonogenic survival as previously described in detail (11). Briefly, each well of 12-well trays was inoculated with 500 to 1,000 cells in 1 mL supplemented medium and incubated overnight to allow attachment and resumption of proliferation. Drug was added, in triplicate for each dose, and incubation continued for 1 hour. Cells were then washed free of residual drug and incubated in fresh, supplemented medium for 5 to 14 days to allow formation of colonies. After staining with 0.5% methylene blue in 1:1 methanol/H₂O (v/v), colonies containing ≥50 cells were counted by light microscopy at 40×. Drug sensitivity was quantitated by linear regression analysis of survival curves (log surviving fraction versus dose; see ref. 11, especially Fig. 1) to obtain the three variables, LD₁₀, D_T and D₃₇. LD₁₀ is the dose required to reduce survival to 10%. D_T, the threshold dose, is the concentration of alkylator tolerated without lethality and is indicated by a shoulder on survival curves. D₃₇ is a measure of the rate of cell killing and is defined by the final slope of the linear portion of the survival curve. Survival curves were determined from three or four independent experiments, as indicated, in which every dose was assayed in triplicate (i.e., 9 or 12 determinations per drug concentration); in some instances, error bars are not visible on individual data points because the SDs were so small.

O⁶-methylguanine-DNA methyltransferase. The MGMT activity of whole cell extracts was assayed by quantitating transfer of radioactivity from a DNA substrate containing [methyl-³H]O⁶-methylguanine to protein, as previously described in detail (18, 19). Extracts were prepared by gently resuspending washed cell pellets (50 μL/10⁶ cells) in 25 mmol/L Tris-HCl (pH 8.0), 5 mmol/L EDTA, 600 mmol/L NaCl, 0.01 mmol/L DTT, 0.2 mmol/L phenylmethylsulfonyl fluoride, 20 μg/mL each of leupeptin and pepstatin, 10% glycerol, and 0.1% NP40. After incubation on ice for 40 minutes, debris was pelleted by centrifugation at 10,000 × g for 30 minutes. Multiple aliquots of supernatant were flash-frozen in liquid nitrogen and stored at −80°C. MGMT activity is the mean of at least five determinations that generally differed by no more than 20%. Validation of the assay and controls indicating that the wide range of MGMT activity observed is unlikely to be due to degradation of MGMT and/or its [³H] DNA substrate during extraction and assay, or to a diffusible inhibitor in extracts, as have been described elsewhere (18, 19).

O⁶-methylguanine-DNA methyltransferase content. MGMT activity was detectable in 15 of the 16 pediatric brain tumor lines, ranging 7-fold from 35 to 249 fmol/10⁶ cells (i.e., ∼21,000-150,00 molecules per cell; Table 2). The medulloblastoma-derived line Res 256 had no detectable activity (<0.25 fmol/10⁶ cells or <151 molecules per cell) and was designated Mer⁻ (i.e., Methyl repair deficient). Mean activity did not differ significantly between the 11 glioma and four Mer⁺ medulloblastoma lines (102 ± 62 versus 126 ± 43 fmol/10⁶ cells). Incubation of the Mer⁺ lines with 20 μmol/L O⁶-benzylguanine eliminated measurable MGMT activity within 1 hour, and activity remained undetectable for at least 72 hours in the presence of the inhibitor (data not shown). Hence, the lines do not harbor mutant MGMT molecules that are insensitive to O⁶-benzylguanine (21, 22, 26).

1,3-Bis(2-chloroethyl)-1-nitrosourea cytotoxicity. The cytotoxicity of BCNU was quantitated by assay of clonogenic survival. The variables LD₁₀, D_T, and D₃₇ that define drug sensitivity were derived from survival curves such as those in Fig. 1 and are compiled in Table 2. The Mer⁺ lines exhibited little variability in resistance, differing in LD₁₀ by <2-fold (59-111 μmol/L). The Mer⁻ line Res 256 was substantially more sensitive (LD₁₀ = 8.8 μmol/L), illustrative of the importance of MGMT as a resistance mechanism. The survival curves for all lines displayed shoulders (e.g., Fig. 1), indicative of either complete repair and/or tolerance of potentially lethal lesions induced at low drug doses. D_T defined by the shoulders in the Mer⁺ lines was about the same as, or greater than, clinically achievable plasma concentrations of BCNU (25-30 μmol/L; ref. 27). Thus, cells survived this BCNU dose without lethality, a finding of potential translational relevance. D₃₇, defined by the slope of the final linear portion of the survival curves, was constant within each line (e.g., Fig. 1), indicative of uniform susceptibility to killing at doses greater than D_T.

To examine the contribution of MGMT to resistance in the Mer⁺ lines, we quantitated the potentiation of BCNU cytotoxicity produced by O⁶-benzylguanine. Cells were incubated with 20 μmol/L O⁶-benzylguanine for 18 to 20 hours before, during, and for 18 to 20 hours after BCNU exposure to ensure ablation of MGMT activity during formation of cytotoxic adducts and induction of cytotoxicity. As exemplified in Fig. 1, O⁶-benzylguanine reduced LD₁₀ an average of 2.6-fold (range, 1.8- to 3.1-fold; Table 2). The changes in LD₁₀ reflected an average 3.3-fold decrease in D_T, indicating that MGMT plays a significant role in resistance to low doses of BCNU; in all Mer⁺ lines, D_T was reduced to a level beneath the clinically achievable plasma concentration of 25 to 30 μmol/L (27). D₃₇ was reduced 2.2-fold on average and was constant in each line, demonstrating the absence of subpopulations of more resistant cells, including cells harboring O⁶-benzylguanine-insensitive MGMT. Extending incubation with O⁶-benzylguanine to 72 hours after BCNU exposure did not further increase cytotoxicity (data not shown). As expected, the BCNU sensitivity of Mer⁻ Res 256 was unchanged by O⁶-benzylguanine (Table 2).

Temozolomide cytotoxicity. Temozolomide sensitivity, determined by clonogenic assay, is illustrated in Fig. 2 and compiled for all lines in Table 3. As in the case of BCNU, the Mer⁺ lines varied little in sensitivity, differing by ∼2-fold in LD₁₀ (327-685 μmol/L). The Mer⁻ line Res 256 displayed much greater sensitivity (LD₁₀ = 12 μmol/L). Each line displayed a shoulder of resistance (D_T) and exhibited constant D₃₇. For all Mer⁺ lines, D_T was greater than the plasma concentration of temozolomide that is clinically achievable (100 μmol/L; refs. 28, 29). As documented in Fig. 2 and Table 3, ablating MGMT activity greatly potentiated temozolomide cytotoxicity, reducing LD₁₀ an average of 26-fold (range, 15- to 46-fold). Lower LD₁₀ reflected large decreases in both D_T and D₃₇. Remarkably, D_T was diminished 42- to 730-fold (average = 138-fold), showing that MGMT was almost exclusively responsible for insensitivity to low doses of temozolomide. Importantly, D_T in the presence of O⁶-benzylguanine was far lower than the clinically achievable plasma level of 100 μmol/L. As for BCNU, extending incubation with O⁶-benzylguanine to 72 hours after temozolomide exposure did not further increase cytotoxicity (data not shown).

O⁶-Methylguanine-DNA methyltransferase activity does not predict alkylator sensitivity or potentiation by O⁶-benzylguanine. The foregoing data show that MGMT is a biochemically demonstrable and frequently predominant mechanism of alkylator resistance in our lines. Nonetheless, linear regression analysis revealed no statistically significant relationship between MGMT activity and LD₁₀ for BCNU (r = 0.344, P ≤ 0.21) or temozolomide (r = 0.304, P ≤ 0.25) in the Mer⁺ lines, indicating that MGMT content is not predictive of alkylator sensitivity. The lack of correlation is exemplified by Res 186 and Res 280, lines that differ 7-fold in MGMT activity but differ little in LD₁₀ for BCNU (88 versus 100 μmol/L) and temozolomide (478 versus 327 μmol/L). In addition, MGMT activity was not predictive of the O⁶-benzylguanine-induced reduction in LD₁₀ for either BCNU (r = 0.344, P ≤ 0.21) or temozolomide (r = 0.022, P ≤ 0.94).

Prolonged incubation with O⁶-benzylguanine is required for maximal potentiation of 1,3-bis(2-chloroethyl)-1-nitrosourea and temozolomide cytotoxicity. Maximal potentiation of BCNU cytotoxicity in adult glioma cell lines requires incubation with O⁶-benzylguanine before and for ∼16 to 24 hours after alkylator exposure (e.g., refs. 26, 30, 31). To ascertain if this is also true for our pediatric lines, we examined the effect of omitting post-alkylator incubation with O⁶-benzylguanine in the medulloblastoma line UW228-1. Importantly, MGMT activity was undetectable at the time alkylator was added. As illustrated in Fig. 3A, LD₁₀ was reduced 3.5-fold (27 versus 79 μmol/L) in cells incubated with 20 μmol/L O⁶-benzylguanine both before and after alkylator exposure but was reduced only 1.3-fold (60 versus 79 μmol/L) if incubation after alkylation was omitted. Comparable results were found for Res 186, Res 196, UW228-2, UW228-3, and UW467 (data not shown), lines with MGMT activities that ranged from 35 to 154 fmol/10⁶ cells (Table 2). The requirement for prolonged incubation with O⁶-benzylguanine after BCNU treatment has been attributed to the slow conversion (∼6-12 hours) of O⁶-chloroethylguanine to the cytotoxic interstrand cross-link 1-(3-cytosinyl), 2-(1-guanyl)ethane (32, 33). If this were the sole basis for the requirement, one would predict that extended incubation with O⁶-benzylguanine after exposure to methylating agents would not be necessary to achieve maximal potentiation because cytotoxic O⁶-methylguanine adducts are formed essentially immediately and the methylating species is short-lived (8, 9). However, as shown in Fig. 3B, omitting post-incubation with O⁶-benzylguanine resulted in no potentiation whatever of killing by MNNG, a methylator that produces an adduct spectrum similar to that of temozolomide (8, 9). Comparable results were observed for Res 186, Res 196, UW228-2, UW228-3, and UW467 treated with temozolomide (data not shown). The requirement for the continued presence of O⁶-benzylguanine has been attributed to a need for persistent O⁶-methylguanine at ongoing replication forks to induce cytotoxicity (9, 34). The foregoing data indicate that resynthesis of MGMT after alkylation can limit (in the case of BCNU) or entirely negate (in the case of temozolomide) the benefit of O⁶-benzylguanine, if the inhibitor is removed from the medium when cells are alkylated.

Having concluded that new MGMT synthesized after alkylation affords partial protection against BCNU killing, we next examined the effect of very low levels of MGMT present at the time of alkylation and afterwards. To approach this question, we exploited the observation that MGMT activity is partially restored within a few hours after O⁶-benzylguanine is removed (31–33). In preparatory work, we examined the time course of resynthesis of MGMT activity in UW228-1 following incubation with 100 μmol/L O⁶-benzylguanine for 24 hours. As shown in Fig. 3C, MGMT activity was detectable within 2 hours after removing O⁶-benzylguanine, reached a plateau of 12% to 15% of untreated control between 24 and 48 hours, and declined thereafter to 2% of control at 120 hours. A similar pattern of repletion has been observed in a human colorectal carcinoma cell line (32, 33). The pattern has been attributed to a complex interaction among (a) synthesis of new MGMT, (b) equilibration of cellular stores of O⁶-benzylguanine between soluble and membranous compartments and with the extracellular medium, and (c) inactivation of newly synthesized MGMT by residual O⁶-benzylguanine (31–33). In independent experiments, we observed that a second medium change 48 hours after the first permitted only a transient 10% increase in activity followed by a gradual decline to 1% of the level of untreated cells within 3 days (data not shown). These findings indicate that a single treatment with 100 μmol/L O⁶-benzylguanine can produce a substantial, albeit incomplete, and persistent reduction of MGMT activity. Having thus established the time course of repletion, we determined the effect on killing of low levels of MGMT present at the time of alkylation. We added alkylating agent at 4, 8, 24, and 120 hours after removing O⁶-benzylguanine when MGMT activity was 4%, 4%, 10%, and 2% of untreated controls, respectively (Fig. 3C). Remarkably, treatment with BCNU or MNNG in the presence of these greatly reduced MGMT levels produced killing indistinguishable from that of control cells that had not been incubated with O⁶-benzylguanine and had a full complement of MGMT (data not shown). Apparently, the presence of low levels of MGMT activity at the time of alkylation and during subsequent incubation can afford complete protection against chloroethylator and methylator killing, at least in our experimental circumstances.

The contribution of MGMT to alkylating agent resistance, as well as the promise and limitations of O⁶-benzylguanine in suppressing resistance, have been thoroughly examined in numerous cell lines and xenografts derived from adult gliomas (11–16). Here we provide an extensive analysis of cell lines derived from pediatric brain tumors, particularly gliomas, for which data are lacking. Our data show that MGMT is a major determinant of BCNU resistance (Table 2) and a predominant determinant of temozolomide resistance (Table 3) in glioma as well as medulloblastoma cells. On average, MGMT-mediated repair of O⁶-alkylguanine lesions was responsible for ∼60% of LD₁₀ for BCNU and ∼95% of LD₁₀ for temozolomide. The importance of MGMT in alkylator resistance is highlighted by the large effect on D_T, the drug dose tolerated without lethality: 70% of D_T for BCNU and ≥95% of D_T for temozolomide reflected the activity of MGMT. In some instances, reduction of D_T was almost exclusively responsible for BCNU sensitization (Table 2). Of potential clinical relevance, ablation of MGMT reduced D_T to values less than or equal to the clinically achievable serum levels of 25 to 30 μmol/L for BCNU (27) and 100 μmol/L for temozolomide (28, 29). MGMT also reduced the rate of killing (i.e., D₃₇) and, on average, was responsible for 65% of D₃₇ for BCNU and >95% of D₃₇ for temozolomide. Despite the important contribution of MGMT to resistance, we found that MGMT activity was not correlated with sensitivity to BCNU or temozolomide, and with O⁶-benzylguanine-mediated potentiation of cytotoxicity. These data emphasize that lack of correlation can be observed even when MGMT is the dominant mechanism of alkylator resistance and highlight the multifactorial nature of resistance.

We found that O⁶-benzylguanine produced markedly greater sensitization to temozolomide than to BCNU. On average, we observed a 10-fold greater reduction in LD₁₀ (26 ± 8- versus 2.6 ± 0.4-fold), a notable 42-fold greater reduction in D_T (138 ± 177-versus 3.3 ± 1.2-fold) and a 6.5-fold greater reduction in D₃₇ (14 ± 5.3-versus 2.2 ± 0.8-fold). This discrepancy in the contribution of MGMT to resistance likely reflects multiple factors, including the preference of MGMT for O⁶-methylguanine (17), the greater proportion of O⁶-alkylguanine adducts produced by temozolomide (7% versus 3.5%; refs. 8, 35), and the absence of alternative mechanisms to repair O⁶-methylguanine. The lesser effect on BCNU cytotoxicity may also reflect the conversion of O⁶-chloroethylguanine to lesions other than interstrand cross-link (e.g., O⁶-hydroxyethylguanine; ref. 6) and the efficient repair of interstrand cross-link (36). We have also observed greater potentiation of temozolomide killing by O⁶-benzylguanine in adult glioma lines, although the differential was not as large (13).

Although MGMT is a major determinant of alkylator sensitivity in our lines, our data indicate that additional mechanisms contribute to resistance. These mechanisms are manifested by the shoulders of resistance (e.g., Figs. 1 and 2), and the heterogeneity of alkylator sensitivity that remain among the Mer⁺ lines after treatment with O⁶-benzylguanine (Tables 2 and 3). The failure of O⁶-benzylguanine to increase the BCNU and temozolomide sensitivity in the Mer⁺ lines to that of Mer⁻ Res 256 also exemplifies the differential operation of additional determinants. Multiplicity of resistance mechanisms is well established. For example, we have shown that repair of abasic sites and N-alkylpurines sensitizes Mer⁻ and Mer⁺ adult glioma cells to BCNU and temozolomide (37).⁶

Resistance may also be mediated by mechanisms that promote recovery of replication forks that have been stalled or disrupted by O⁶-alkylguanine lesions. We have observed that suppressing the Werner syndrome protein, a DNA helicase believed to contribute to the resolution of stalled replication forks (38), sensitizes adult glioma cells to temozolomide (39) and BCNU.⁷ Importantly, sensitization occurred only in the absence of MGMT, implying that persistent O⁶-alkylguanine adducts cause lethal events at replication forks. Yet other mechanisms (e.g., perturbation of cell cycle arrest or cell death), not all involving repair or tolerance of alkyl lesions, may promote alkylator resistance (10, 40, 41). Characterization of additional determinants of resistance will be central to developing new targets for antiresistance therapies in tumors that become insensitive to O⁶-benzylguanine by selection for MGMT variants (e.g., ref. 26) or loss of mismatch repair (9).

Our results show that maximal suppression of chloroethylator and methylator resistance requires elimination of detectable MGMT activity (<0.25 fmol/10⁶ cells or <150 molecules per cell) with O⁶-benzylguanine not only before and during but also for a prolonged period after alkylator exposure. Previous work has shown that potentiation of BCNU cytotoxicity in human colorectal carcinoma and adult glioma cells is reduced if incubation with O⁶-benzylguanine after alkylator exposure is omitted (30–34). In accord, we also observed a diminished effect on the potentiation of BCNU killing if O⁶-benzylguanine post-treatment was omitted (e.g., Fig. 3A). In contrast, omitting post-treatment resulted in no potentiation whatever of MNNG or temozolomide killing (e.g., Fig. 3B). We note that there was no detectable MGMT in cells at the time of alkylation. Comparable results have been found for temozolomide killing in an adult glioma line (40). These data indicate that small amounts of MGMT synthesized after alkylation reduce the abundance of O⁶-methylguanine below the level necessary to induce cytotoxicity. Persistent O⁶-methylguanine adducts are believed to induce cytotoxicity mediated by mismatch repair during rounds of DNA replication that follow alkylator exposure (9). That MGMT does not provide full protection against BCNU when post-treatment is omitted probably reflects the fact that O⁶-chloroethylguanine is a relatively poor substrate for MGMT, as well as the slow formation of cytotoxic interstrand cross-links.

We found that very low levels of MGMT activity (2-10%), if present at the time of alkylation and after, provide full protection against both BCNU and MNNG. This result differs from our earlier findings that subtotal (2- to 3-fold) suppression of Ape1/Ref-1 activity (37) and of the Werner syndrome protein (39) decreased resistance to BCNU and temozolomide. It is not readily apparent why the observed ≥90% reduction of MGMT activity at the time of and after alkylator exposure does not enhance lethality. Conceivably, low levels of MGMT may promote full resistance if activity is preferentially directed to O⁶-alkylguanine lesions in the path of replication forks and/or to O⁶-methylguanine mispairs caught in repetitive cycles of mismatch repair (9).

The poor prognosis for primary brain tumors in children reflects, in part, the lack of effective therapies for recurrent malignant tumors and for local control of partially resected low-grade gliomas (1–3). Clinical trials have shown the efficacy of temozolomide against newly diagnosed and recurrent adult malignant gliomas (42), and emerging evidence indicates that temozolomide may be effective in treating recurrent low-grade adult gliomas (43). These findings, together with the lower probability of cumulative marrow toxicity associated with temozolomide (42), have stimulated interest in this methylator to treat newly diagnosed and recurrent pediatric brain tumors. Recent pilot studies in children have revealed promising activity of temozolomide against medulloblastoma and brain stem tumors (44, 45). Results in high-grade gliomas, however, have been mixed (45, 46). The data we present here suggest that O⁶-benzylguanine could appreciably increase the clinical efficacy of temozolomide in pediatric gliomas and medulloblastomas. A clinical trial to evaluate this premise is ongoing (PBTC-005; http://www.cancer.gov).

Emerging data suggest that the full benefit of suppressing MGMT will require continuous infusion with O⁶-benzylguanine for prolonged periods after alkylator exposure, as evidenced by a recent report that 100 mg/m² O⁶-benzylguanine given 1 hour before BCNU failed to produce any objective responses in adult malignant gliomas (47). A subsequent study found that MGMT activity was detectable in 55% of adult gliomas resected 18 hours after a single infusion of 120 mg/m² O⁶-benzylguanine (48). This work suggests that a single infusion of O⁶-benzylguanine may not provide the prolonged and complete ablation of MGMT activity associated with full sensitization in vitro. Erickson and colleagues have shown that continuous infusion of low levels of O⁶-benzylguanine preceding a larger single bolus can produce near total depletion of MGMT activity in xenografts of the adult glioma line SF767 that persisted for at least 24 hours (30). A similar approach may be clinically efficacious.