O⁶-Benzylguanine: A Clinical Trial Establishing the Biochemical Modulatory Dose in Tumor Tissue for Alkyltransferase-directed DNA Repair¹

Although nitrosoureas and related alkylating agents have a broad spectrum of antitumor activity, resistance develops rapidly. DNA repair enzyme activity mediates resistance to the drug BCNU.³ The cytotoxic action of BCNU occurs through the formation of alkyl DNA adducts at the O⁶ position of guanine. BCNU, in the presence of an oxygen or nitrogen nucleophile, forms a chloroethyl diazonium ion that is highly reactive with the nucleophile. The initial adduct formed, O⁶-(2-chloroethyl) guanine, is unstable and undergoes an intramolecular cyclization forming an O⁶N¹-ethanoguanine intermediate, which reacts with cytosine on the opposite strand, forming an N¹-guanine, N³-cytosine interstrand cross-link (1, 2, 3). The process occurs over a period of 1–12 h. These cross-links disrupt DNA synthesis and cause chromosomal aberrations, rearrangements, sister chromatid exchanges, and strand breaks (4, 5), each of which correlates with toxicity (6). As few as 10 cross-links are cytotoxic to tumor cells (7). Repairing adducts, or preventing cross-link formation by repairing the O⁶N¹-ethanoguanine intramolecular intermediate for 18 h after BCNU therapy, impairs the cytotoxic action of the drug (8, 9).

AGT is a DNA repair enzyme (homologous to the Escherichia coli ada alkyltransferase) that repairs adducts at the O⁶ position of guanine. Each AGT molecule removes one adduct through covalent binding of the alkyl group to the cysteine residue at amino acid number 145. During this process, irreversible inactivation of the protein occurs, and synthesis of new molecules is required to regenerate AGT activity (10, 11). The natural targets for AGT are small alkyl groups including methyl, ethyl, isopropyl, hydroxyethyl, and chloroethyl adducts, although the O⁶ bond of the O⁶N¹-ethanoguanine intermediate is also a target (9). The N¹-guanine, N³-cytosine cross-link, once formed, is not a substrate for AGT (1). The “suicide protein” properties of AGT make it a unique target for biochemical modulation.

In vitro and human xenograft studies have shown a close correlation between AGT depletion and alkylating agent activity (12). AGT depletion for 12–18 h is needed for the enhancement of response (7, 13). Increased AGT activity is found in many human solid tumors including colon cancer (14), malignant melanoma (15), breast cancer (16), gliomas and medulloblastomas (17, 18), lung cancer, rhabdomyosarcomas (19), and acute myeloid leukemias (20). Malignant tumor cell lines with high levels of AGT activity are resistant to alkylating agents (21). Human tumor xenografts with high AGT activity are similarly resistant to BCNU, whereas those with low AGT activity are sensitive (22, 23). Sensitive tumors transduced with the methylguanine methyltransferase gene, which encodes the AGT enzyme, acquire resistance to BCNU (24). Modulation of AGT activity with O⁶-methylguanine has been demonstrated, but depletion of AGT was incomplete (20). O⁶-Methylguanine combined with BCNU was unsuccessful in increasing the therapeutic index of BCNU in human tumor xenografts (25). Streptozotocin also depletes AGT activity (incompletely), but the streptozotocin/BCNU combination causes severe hematological toxicity (26).

BG is a potent AGT-inactivating agent (27). It binds to the same cysteine residue on the AGT molecule that is used for alkyl group transfer and inactivates the enzyme stoichiometrically (10). Transfer of the benzyl group to the alkyltransferase protein causes a marked decrease in stability of the protein in HT29 cells (28). Nontoxic doses of BG render tumor cells more sensitive to BCNU; the greatest enhancement occurs in cells with high AGT activity (29). BG levels >200 ng/ml for 12–18 h after treatment with BCNU is required for depletion of AGT activity to undetectable levels in vitro. Newly synthesized AGT binds preferentially to DNA adducts, limiting the potential for cytotoxicity unless AGT is depleted to undetectable levels (23, 30). In human tumor xenograft models, sequential BG and BCNU administration showed significant tumor inhibition (31). The combination was superior to the maximal tolerated dose of single-agent BCNU (22). Preclinical toxicology studies of sequential BG/BCNU in mice and dogs showed dose-limiting myelotoxicity. At therapeutic doses, there was no increase in nonhematological toxicity (32, 33). Because mouse AGT is highly resistant to inactivation, the combination in humans may have a wider therapeutic-to-toxic window.

Limited water solubility necessitates BG administration in 40% polyethylene glycol 400 (34). In rodents, BG disappears rapidly from the circulation (t_1/2 in rats, 1.6 h). Renal excretion of the parent drug accounts for only 8% of the administered dose, suggesting extensive metabolism. The major urinary metabolites in the rat are debenzylated guanine, 8-oxo-BG, N²-acetyl-benzylguanine, and N²-acetyl-8-oxo-benzylguanine. Thirty-seven % of the parent drug is converted to 8-oxo-BG in rats. In mice and nonhuman primates, N²-acetyl metabolites are not found (35, 36). BG distribution studies show rapid uptake by cells, and conversion to 8-oxo-BG occurs through the action of cytochrome P450 isoforms, CYP1A2 (predominantly) and CYP3A4. The cytosolic enzyme aldehyde oxidase also contributes to the oxidation of BG as determined by allopurinol and menadione inhibition studies. 8-oxo-BG has biologically activity that approximates that of BG (35).

To define the clinical role of BG as a modulator of AGT activity in tumor tissue, we performed a Phase I dose escalation study. At our institution, we have used posttreatment tumor tissue biopsy specimens to correlate clinical efficacy with events at a cellular level (26). In this study, we designed a novel paired biopsy system as a dose titration end point for the drug BG. Because BG has no cytotoxic activity, it was combined with BCNU. The objectives of the study were to determine the biochemical modulatory dose of BG for the depletion of AGT activity to undetectable levels in tumor tissue, the dose-limiting toxicity of BG, and to correlate tumor tissue AGT depletion with plasma concentrations of BG and its major metabolites.

Patients enrolled in this study were males or females 18 years of age or older, with metastatic solid tumors, performance status Eastern Cooperative Oncology Group 2 or better, with adequate renal (creatinine <1.6) and hepatic function (AST < twice upper limit of normal). All patients were required to have clinically or radiographically measurable lesions, accessible to biopsy. Institutional Review Board approval and patient informed consent was obtained for pretreatment and posttreatment CT-guided biopsies.

This clinical trial was designed to determine the biochemical modulatory dose of BG that depletes AGT activity to undetectable levels in tumor tissue (BMD^T). In week one, patients received a 1-h infusion of BG alone. Apart from the first three patients (not required by protocol), all patients underwent pretreatment biopsies within 4 days prior to BG and posttreatment biopsies at 2 h (patients 4, 5, and 6) or 18 h (remaining patients) after BG. PBMCs isolated from whole-blood samples before, during (15, 30, and 45 min), and after BG (at end of the infusion designated as time point 0 and at 2, 6, 18, and 24 h after the end of the infusion) were analyzed for AGT activity. The plasma concentrations of BG and its major metabolites were assessed in plasma samples drawn before, during, and at selected times during the 24 h following the end of the BG infusion. On day 15, BG administration was followed 1 h later by a 1-h infusion of 13 mg/m² BCNU. AGT activity in PBMC and samples for BG/BG metabolite concentrations were drawn at the same time points as on day 1 to assess the effect of BCNU on the depletion of AGT activity in PBMCs and on the plasma concentrations of BG and its metabolites. Tumor tissue biopsies were not performed on day 15. The dose of BG was increased in cohorts of three patients according to a modified Fibonacci dose escalation plan that initially escalated the BG dose by 33% increments from 10.0 to 36.0 mg/m². A subsequent protocol revision allowed 50% dose increments of BG from 36.0 to 54.0, 80, and 120 mg/m².

More than 100 clinical tumor biopsies have been performed to date at our institution, and this technique has proven to be safe. The solid tumor samples were obtained by CT-guided or percutaneous cutting needle biopsies and were frozen immediately on-site in liquid nitrogen. The tissue, 2 × 2 × 15 mm, was divided into six to eight sections alternately assigned for histological examination and AGT biochemical assay. Each segment of the biopsy selected for AGT biochemical assay contained 5–20 mg of tissue (wet weight) and 0.4–1.0 mg of protein. The AGT assay requires the use of as little as 2–5 mg wet weight of tissue. Tumor biopsies contain mostly tumor but may contain areas of necrosis, fibrosis, or normal tissue. Pre- and posttreatment values represent mean AGT activity from duplicate assays in each segment of the biopsy. Two or more (up to five) segments were assayed for AGT activity from each biopsy, resulting in between 4 and 10 values for each biopsy (Table 2).

The activity of alkyltransferase in tumor tissue extracts was measured as removal of the [³H]methyl adduct from the O⁶ position of guanine in DNA by incubating cell or tissue extracts with DNA substrate (37). Briefly, substrate [³H]methyl-DNA was prepared by reacting calf thymus DNA with [³H]methylnitrosourea (specific activity was 0.06 fmol/dpm, corrected for decay of [³H]) as described (37). New substrate was also directly compared with older substrate so that the reporting of AGT activity from this laboratory has remained constant over the past 15 years. The specific activity was confirmed by establishing a standard curve using calf thymus DNA as a negative control and a NIH3T3-methylguanine methyltransferase transfectant cell extract as a constant substrate. During method development, we determined that removal of O⁶-methylguanine is linear with respect to N⁷-methylguanine when increasing amounts of AGT containing cell extract is added in the range of 12 to 85% removal of O⁶-methylguanine. All assays were performed within this range.

Each tumor segment (about 2 × 3 mm) was placed in 250 μl of cell extract buffer [CEB containing 70 mm HEPES (pH 7.8), 0.1 mm EDTA, 5% glycerol, 1 mm DTT, and 25 μm spermidine] and sonicated three times for 10 s at 4°C to complete cell disruption using a microsonicator equipped with a 3/32” diameter probe (Branson Ultrasonics Corp., Danbury, CT). Aliquots (20 μl) for DNA quantification by Hoescht dye stain (37) were removed, and tumor tissue extracts were centrifuged at 13,000 rpm for 2 min to remove cellular debris. Protein concentration was determined by Bio-Rad protein assay.

Fifty to 250 μg protein of tumor tissue extract was incubated with [³H]methyl DNA in CEB buffer in the total volume of 300 μl for 60 min at 37°C. The reaction mixture contained excess substrate DNA such that the total AGT activity could be determined. The reaction was stopped with 7.5% trichloroacetic acid at 4°C for 30 min. The precipitate was collected by centrifugation at 13,000 rpm for 2 min and washed with 300 μl of 80% ethanol. Methylated purines were liberated from precipitated DNA during hydrolysis with 150 μl of 0.1 n HCl at 80°C for 1 h. [³H]O⁶-methylguanine and [³H]N⁷-methylguanine (which was constant during the incubation and served as internal standard) in supernatant were separated by reverse-phase HPLC and quantitated by liquid scintillation counting. Radioactivity (dpm) in each peak (N⁷-methylguanine and N⁶-methylguanine) derived from the substrate ranged from 350 to 800 dpm for the O⁶-methylguanine peak (21–48 fmol) and 12-fold higher values for N⁷-methylguanine. Peak values were corrected for background radioactivity based on both the positive and negative controls. AGT activity was expressed as fmol O⁶-methylguanine removed/μg DNA. Duplicate assays were performed on each biopsy segment.

The limit of detection is based on the linear portion of the assay and was defined as AGT activity (fmol/μg DNA) corresponding to 12% of the O⁶-methylguanine present in the substrate [³H]methyl DNA used in each assay. For tumor biopsy samples, the limit of detection value was 0.1 fmol/μg of sample DNA.

PBMCs from whole-blood samples were obtained by Ficoll-Hypaque separation and assayed for AGT as described above. For PBMCs, the limit of detection value was 0.05 fmol/μg of sample DNA and is lower than for tumor samples because more cell extract, and thus more [³H]DNA substrate, is used in each assay.

Assays for BG and 8-oxo-BG have been developed at our institution (38, 39). Samples of whole blood are collected in EDTA; plasma is harvested, flash frozen, and stored for assay. After extraction in NaOH, BG concentration is measured by HPLC with a reverse phase C₁₈ column using an ammonium acetate/methanol buffer. O⁶-Chlorobenzylguanine was used as an internal standard. BG is detected by UV absorption with a limit of detection of 12.5 nanogram/ml of serum. A single metabolite peak cochromatographed with synthetic 8-oxo-BG under three sets of HPLC conditions. The UV spectrum of the peak matched that of synthetic 8-oxo-BG, and HPLC-mass spectrometry gave evidence consistent with peak identity as 8-oxo-BG.

A total of 30 patients were treated with BG in this study, 19 females and 11 males. The mean age was 58 years. Seventeen patients had metastatic colorectal carcinomas, and the remaining 13 patients had metastatic breast cancer, carcinoma of the lung, gastric cancer, carcinoma of unknown primary, carcinoma of the pancreas, renal cell carcinoma, and carcinoma of the maxillary sinus. Patients enrolled in this study received BG in doses ranging from 10 to 120 mg/m² (Table 1).

Cutting needle biopsies using local anesthesia, done under CT scan guidance (21 patients) or by percutaneous biopsy of palpable masses (6 patients), were an integral part of the dose escalation strategy. Accordingly, the feasibility of this technique was part of the assessment in this study. Seventeen of the 21 CT-guided biopsies were from hepatic metastatic deposits, 2 from abdominal masses, 1 from lung, and 1 from renal cell carcinoma encasing the spleen. Six of the patients underwent percutaneous biopsies (Table 1). Irrespective of location, tandem biopsies less than 48 h apart were performed safely. Some patients experienced transient pain at the biopsy site. One patient had a subcapsular hemorrhage of the liver, which did not require further intervention but which precluded a second biopsy. No other serious sequelae were observed. One patient refused the second biopsy. In five patients, either one or both biopsies contained insufficient viable tissue (usually due to necrotic tumor) for analysis. By targeting the outer rim of tumor deposits, rather than the central portion of the metastasis, necrotic areas of tumor were mostly avoided. Evaluable paired biopsies containing tumor were obtained in 20 of 27 patients. The technique of immediate freezing in liquid nitrogen resulted in satisfactory tissue preservation for AGT assay. Results were reproducible across all tumor types and biopsy locations.

The depletion of AGT activity to undetectable levels in tumor tissue biopsies was used to define the optimal BG dose. Pretreatment AGT activity showed variability between patients with the same tumor type and between tumor types (Table 2). All tumors had detectable AGT activity pretreatment. Mean baseline levels of AGT activity were 17.8 fmol/μg DNA (range, 1.7–49.5 fmol/μg DNA).

In the cohort of patients who had biopsies performed 2 h after 10.0 mg/m² BG, mean AGT activity postreatment was 1.3 fmol/μg DNA (range 0.5–2.1). In patients treated with 10 mg/m² BG who underwent tumor biopsies 18 h after BG infusion, mean residual activity was 1.3 fmol/μg DNA (range, 1.2–1.4). As the dose of BG was increased, residual posttreatment activity persisted. At a BG dose of 54 mg/m², AGT activity was undetectable in the biopsy segments from one of three patients and in two of three biopsy segments from a second patient (Table 2). At 80 mg/m², residual AGT activity was seen in three of four biopsy segments from one patient, and therefore the BG dose was escalated without treating additional patients at this dose. With 120 mg/m² BG, no detectable AGT activity was present in tumor tissue biopsy segments from three evaluable patients. The mean variance for all duplicate segment AGT assays was 9.3%, indicating the highly reproducible nature of the tumor AGT microassay.

Fig. 1 shows pre- and posttreatment biopsies from patient 30 (renal cell carcinoma). The bar in the upper panel of the figure represents the pretreatment needle biopsy core of tissue cut into segments labeled A–H. Segments A, C, and F were used for histological assessment (only segments C and F are shown) to ensure that tumor tissue is present, and that intervening segments assessed for biochemical AGT activity truly represent tumor values. In this patient, segments B and E were analyzed for biochemical AGT activity. Mean pretreatment AGT activity was 8.7 and 4.4 fmol/μg DNA (duplicate assays) for segments B and E, respectively. Segments D, G, and H were used for histological and immunohistochemical evaluation. Similarly, the lower panel shows segments A–H for the post-BG biopsy. Here, AGT activity is not detectable in the segments analyzed for biochemical AGT activity (B and E). The histological segments A, C, D, and H (histology for segments C and H only are shown) confirm the presence of tumor tissue flanking the segments in which AGT biochemical assays were performed.

Fig. 2 shows a similar profile of the biopsy for patient 26 (colorectal carcinoma). In the upper panel, segments B and E were assayed for AGT activity. Mean AGT activity for segments B and E were 26.0 and 28.0 fmol/μg DNA, respectively (duplicate assays). Segments C and F show representative histology confirming the presence of tumor tissue flanking the segments assayed biochemically. The lower panel shows the posttreatment biopsy where segments B and E show no detectable AGT activity, and segments C and H are shown as representative histological confirmation.

Pretreatment AGT activity in PBMC did not correlate with tumor tissue AGT activity (Fig. 3). Mean pretreatment AGT activity was 8.7 fmol/μg DNA and showed less intersubject variation (range, 4.2–15.1) than the tumor biopsies. During BG infusion at 10 mg/m² (the lowest dose tested), mean AGT activity fell from 8.0 fmol/μg DNA to 1.8 fmol/μg DNA within the first 15 min. A nadir of 0.4 fmol/μg DNA was observed at 18 h after infusion, and sustained (but incomplete) inactivation (0.8 fmol/μg DNA) was seen at 24 h. With BG dose escalation, similar patterns of AGT activity in PBMCs were observed. At a BG dose of 54 mg/m², AGT activity was undetectable at the end of the 60-min BG infusion, but residual activity was seen at both 18 and 24 h following the end of the infusion. At the 120 mg/m² dose, greater depletion of AGT activity was seen at 24 h, where mean AGT activity was 0.19 fmol/μg DNA compared with mean residual activity of 0.73 fmol/μg DNA at 54 mg/m² (Fig. 4).

AGT recovery at day 8 was incomplete in all dosage groups, and the extent of recovery did not correlate with dose (Fig. 4). By day 15, AGT activity in most (but not all) patients had recovered to baseline levels. The extent of recovery did not appear to be dose related. In week 3, the BG/BCNU combination did not alter the pattern of depletion of AGT activity in PBMCs. The same temporal pattern of recovery was observed in samples taken on days 22 and 29, 7 and 14 days, respectively, after the BG/BCNU sequential dosing (data not shown).

During the infusion of lower doses of BG, there was a rapid approach to plasma steady state. With BG dose escalation, a dose-dependent increase in the Cmax of BG was observed (data not shown). Steady state was not reached, even at the highest BG dose tested (data shown for the 120 mg/m² dose; Fig. 5,A). At the end of the infusion (time point, 0 min), rapid elimination of BG from the plasma occurred. Two h after the end of the infusion, BG was undetectable in plasma in dosage groups <54 mg/m². In the patients treated with the 120 mg/m² dose, BG plasma concentrations were undetectable 6 h after the end of the infusion (Fig. 5 A). The data are consistent with a two-compartment model, comprising a rapid elimination process with a t_1/2 of 0.1 h and a slower elimination process with a t_1/2 of about 0.6 h. Elimination processes were not saturated in doses up to 120 mg/m².

During week 1, within 15 min of commencing the BG infusion, there was a rise in plasma 8-oxo-BG (the major metabolite of BG). Mean peak plasma concentration (Cmax) of 5940.3 ng/ml (SD, 1447.9) was seen 90 min after the end of the BG infusion (Fig. 5,B). Mean 8-oxo-BG plasma concentrations 12 h after infusion were 1990.4 ng/ml (SD, 1137.3), and at 24 h after infusion, mean 8-oxo-BG concentrations were 431.1 ng/ml (SD, 733.2). Plasma 8-oxo-BG concentrations >200 ng/ml at 24 h were found in two of the three patients whose tumor tissue had undetectable AGT activity (patients 28 and 30). The 24-h concentrations were 1733.6 and 251.4 ng/ml, respectively. The third patient (no. 26) with undetectable tumor tissue AGT activity had 12- and 24-h 8-oxo-BG concentrations of 1201 and 72.6 ng/ml, respectively. Because of limitations on the amount of blood that could be taken from each patient, plasma concentrations of 8-oxo-BG were measured at 18 h in only one patient (no. 29), and this value was 278.3 ng/ml. 8-oxo-BG has a t_1/2 of ∼3.6 h. BCNU given 1 h after BG in week 3 had no impact on BG/8-oxo-BG plasma concentrations (Fig. 5).

BG given alone (in week 1) was well tolerated. No dose-limiting toxicity was observed. Transient lymphocytopenia (non dose-limiting) was seen in 12 of the 20 patients during either the first or subsequent cycles of therapy. There were no apparent clinical sequelae associated with this transient lymphocytopenia. All other toxicities observed were due to preexisting conditions or were related to disease progression. BG in combination with 13 mg/m² BCNU was similarly well tolerated. BCNU had no impact on the transient lymphocytopenia observed with BG.

In other trials, AGT activity in PBMCs have been used to establish the clinical dose of BG (40). The major paradigm of this study was the demonstration that tumor tissue biopsy measurement of AGT could be used as a dose-escalation end point for the development of a new anticancer agent. Preclinical (in vitro) models from one of our laboratories (S. L. G.) have shown previously that 90% depletion of AGT in tumor cells is readily achievable with BG, but that depletion of AGT activity to undetectable levels requires doses up to 1 log higher. These in vitro studies further showed that resynthesis of AGT occurs unless BG presence is maintained, and that concentrations of BG >200 ng/ml are needed to maintain AGT activity depletion at undetectable levels. Furthermore, optimal BCNU effect was obtained when AGT activity was depleted to undetectable levels. On the basis of these preclinical studies, we predicted that a similar 1-log dose titration might be needed to deplete AGT activity to undetectable levels in tumor tissue biopsies in our clinical trial.

We have shown previously that CT-guided cutting needle biopsies can be performed safely (41, 42). As has been shown previously in human tumors (15), we observed a wide variation in baseline AGT activity in our series of solid tumors. AGT activity was detected at baseline in all tumor biopsies, unlike primary brain tumors, where as many as 20% of tumors are reported to lack AGT (43). At the lowest tested BG dose (10 mg/m²), >87% depletion AGT activity was seen in tumor biopsies performed 2 h following the end of the BG infusion. Because the critical time for development of DNA cross-links is up to 12–14 h after BCNU administration, all subsequent patients underwent biopsies 18 h after the end of the BG infusion. As the dose of BG was escalated from 10 to 120 mg/m², residual AGT activity remained in posttreatment tumor biopsy specimens from all doses tested, except at the 120 mg/m² dose. BG depleted AGT activity irrespective of the initial AGT level, tumor type, or site of biopsy.

Because tissue concentrations of BG could not be measured, we performed detailed kinetic studies in our trial in an attempt to correlate tissue AGT depletion with plasma concentrations of BG. The kinetic studies from our initial patients treated at 10 mg/m² showed that BG disappeared rapidly from plasma, and that 2 h after the end of the infusion, plasma concentrations were undetectable. As the dose of BG was escalated, the same rapid disappearance of BG was demonstrated, and this rate of disappearance was dose independent. At the highest dose tested (120 mg/m²), BG plasma concentrations were low 2 h after the end of the infusion and undetectable thereafter. We therefore analyzed samples for metabolites of BG. Distinct from metabolism in the rat (35, 36), the major metabolite found was 8-oxo-BG. We confirmed the longer half-life of 8-oxo-BG seen in animals. Readily detectable plasma concentrations were found 24 h after the end of the infusion in all 5 patients treated at the 120 mg/m² dose. We hypothesized that the initial depletion of AGT activity might be due to BG, and that AGT activity might be undetectable (or nearly so) immediately after BG infusion. The tumor tissue biopsies obtained at 2 h post-BG and the PBMC data obtained during, immediately after (0 time point), and at 2 and 6 h following the end of the BG infusion support this concept. We further hypothesized that the AGT activity present at 18 h in both tumor tissue and PBMCs and at 24 h in PBMCs was due to the synthesis of new AGT molecules. It therefore seemed plausible that at higher doses of BG, tissue concentrations of 8-oxo-BG might be sufficiently high at 18 h to result in depletion of AGT activity to undetectable levels. Our data show that 8-oxo-BG concentrations were undetectable at the 18-h time point in all dosage groups from 10 to 54 mg/m². In the 80- and 120-mg/m² dosage groups in patients in whom AGT activity was undetectable in tumor tissue at 18 h, measured or interpolated 8-oxo-BG concentrations were well above 200 ng/ml. We believe that these data are entirely consistent with the preclinical models described previously (32).

Although in the initial phases of our trial the PBMC findings proved useful in confirming the clinical activity of BG, we found that AGT activity in PBMCs was not predictive of the depletion of AGT activity in tumor tissue. If we had used the early (0–6 h) depletion of AGT activity in PBMCs as the dose escalation end point, this would have lead to an underestimation of the BG dose required for depletion of AGT activity to undetectable levels in tumor tissue. The depletion and reappearance of AGT activity in PBMCs appear to be due to new synthesis of AGT molecules. Small but consistent increases in AGT activity were observed from 18 to 24 h postinfusion, suggesting that the rate of resynthesis is slow, although based on preclinical modeling, it may be much greater in tumor tissue (14, 47). The incomplete recovery of AGT activity in PBMCs at day 8 and day 22 in our study is consistent with this hypothesis and with observations in other studies (44, 45, 46). Data from one of our laboratories suggest that resynthesis of AGT in lymphocytes is slower than that observed in tumor cells (47), further suggesting that AGT activity in PBMCs may predict poorly for tumor cell responsiveness to BCNU.

BG administration was well tolerated. Transient lymphocytopenia was the only drug related side effect seen, was not dose-limiting, and no clinical sequelae resulted. We saw no evidence of myelosuppression with BG alone, regardless of dose. Because we used the BMD^T as the primary end point for our study, we were unable to establish the dose-limiting toxicity of BG.

The combination of 120 mg/m² BG plus 13 mg/m² BCNU was similarly safe. Preclinical toxicology studies predicted that myelosuppression might be dose limiting for the BG/BCNU combination. The starting dose of BCNU (13 mg/m²) was chosen to avoid myelotoxicity during the dose escalation of BG. We observed neither myelosuppression nor nonhematological dose-limiting toxicity in any of the 30 patients treated. In an ongoing study, the dose of BCNU is being increased to determine the dose-limiting toxicity and maximum tolerated dose of BCNU given following 120 mg/m² BG.

Understanding the pharmacodynamic events at a cellular level is crucial to the development of strategies to overcome the development of resistance to chemotherapy by tumor cells. Because AGT appears to play a crucial role in the development of resistance to alkylating agents other than BCNU (48), BG may have broad application as an AGT-depleting agent to enhance the efficacy of chemotherapeutic alkylating agents that produce cytotoxic O⁶-alkylguanine adducts. We have shown proof of the concept that in a clinical setting, AGT activity can be reproducibly measured in tumor tissue and PBMCs, and that AGT activity can be depleted to undetectable levels at nontoxic BG doses. From our tumor tissue biopsy data, we have established that a BG dose of 120 mg/m² infused over 1 h should be used in Phase II clinical trials.