Formation and Antitumor Activity of PNU-159682, A Major Metabolite of Nemorubicin in Human Liver Microsomes Free

The clinical use of classical anthracyclines, such doxorubicin and daunorubicin, is limited by their dose-related cardiotoxicity, myelosuppression, and intrinsic or acquired tumor resistance. Among the several anthracycline analogues synthesized in an effort to obtain less toxic and more effective antitumor agents, nemorubicin (3′-deamino-3′-[2″(S)-methoxy-4″-morpholinyl]doxorubicin; MMDX), a doxorubicin derivative bearing a methoxymorpholinyl group at position 3′ of the sugar moiety, has been selected for clinical evaluation. Unlike doxorubicin, MMDX is also highly cytotoxic to a variety of tumor cell lines presenting a multidrug-resistant phenotype both in vitro and in vivo (1–3), and is not cardiotoxic at therapeutic doses (4); it is presently undergoing a phase II/III clinical trial for intrahepatic artery chemotherapy of hepatocellular carcinoma, with a preliminary response rate of about 25%. (5). MMDX is remarkably more potent in vivo than currently clinically available anthracyclines, its maximum tolerated dose being 50- to 130-fold lower than that of doxorubicin, in both experimental animals and humans (1, 6). By contrast, MMDX is only 2- to 10-fold more cytotoxic than doxorubicin in vitro (4, 7).

Previous in vitro studies showed that the cytotoxicity of MMDX is markedly enhanced upon its exposure to NADPH-supplemented liver microsomes from different animal species, including humans, and identified cytochrome P450s (CYP) belonging to the 3A subfamily as the enzymes responsible for hepatic bioactivation of the drug (8–10). Further studies in tumor-bearing and healthy mice revealed that a metabolite(s) formed via CYP3A mediates both therapeutic efficacy and host toxicity of MMDX (9).

To date, the only metabolite of MMDX consistently found in the plasma of patients and experimental animals receiving the drug is 13(S)-dihydro-MMDX, which is formed via stereoselective reduction of the side chain C₁₃-carbonyl group of MMDX (6, 11). Like most of 13-dihydro metabolites of other anthracyclines (12), 13-dihydro-MMDX shows a lower cytotoxicity than its parent compound in vitro.⁵

Three oxidative metabolites of MMDX, i.e., MMDX N-oxide, 3′-deamino-3′-[2″(S)-methoxy-5″-oxo-4″-morpholinyl]doxorubicin (PNU-159696), and 3′-deamino-3″,4′-anhydro-[2″(S)-methoxy-3″(R)-oxy-4″-morpholinyl]doxorubicin (PNU-159682), have been recently isolated from an incubate of the drug with NADPH and liver microsomes from dexamethasone-induced male rats (14). Preliminary cytotoxicity experiments using cultured murine L1210 tumor cells as targets revealed that, whereas MMDX N-oxide and PNU-159696 possess a potency comparable to that of MMDX, PNU-159682 represents a bioactivation product being more than 3,000-fold more cytotoxic than its parent compound (14). Furthermore, although the low therapeutic dosages of MMDX hinder the detection of this drug and its metabolites in biological fluids and tissues, traces of PNU-159682 have been detected in a plasma sample from a cancer patient receiving prolonged i.v. infusion of MMDX, as well as in the liver extracts of tumor-bearing mice receiving optimal antitumor doses of the drug as i.v. bolus.⁶

[¹⁴C]MMDX (1.79 GBq/mmol; radiochemical purity of 93%), unlabeled MMDX, racemic 13-dihydro-MMDX, doxorubicin (all as hydrochloride salts), MMDX N-oxide, and PNU-159682 (as free base), were synthesized at Pharmacia Italia S.p.A. (Nerviano, Italy). Methanol, acetonitrile, ethanol and dichloromethane [all high-performance liquid chromatography (HPLC) grade] were from Carlo Erba Reagenti (Milan, Italy). Oxidized nifedipine [2,6-dimethyl-4-(2′-nitrophenyl)-3,5-pyridinedicarboxylic acid dimethyl ester] was obtained from Ultrafine (UFC, Ltd., Manchester, United Kingdom). Unless indicated otherwise, all other chemicals were purchased from Sigma-Aldrich Co. (St. Louis, MO).

Inhibitory monoclonal antibodies (MAb) to human CYP enzymes were generously donated by Dr. H.V. Gelboin (Laboratory of Molecular Carcinogenesis, National Cancer Institute, NIH, Bethesda, MD). These were: MAbs 26-7-5 (anti-CYP1A2), 49-10-20 (anti-CYP2B6), 281-1-1 (anti-CYP2C8), 763-15-5 (anti-CYP2C9), 1-7-4-8 (anti-CYP2C19), 512-1-8 (anti-CYP2D6), and 3-29-9 (anti-CYP3A4/5; refs. 15–19). Pooled HLMs (mixed gender; n = 21) and microsomes prepared from 10 different human livers were provided by BD Gentest (Woburn, MA). Microsomes from cDNA-transfected human B-lymphoblastoid cells expressing CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9*1, CYP2C19, CYP2D6*1, CYP2E1, CYP3A4, and microsomes from baculovirus-infected insect cells expressing CYP3A5 were also obtained from BD Gentest.

Nifedipine oxidase activity of each individual liver microsomal sample was estimated as described by Guengerich et al. (20) 

Erythromycin N-demethylase activity was determined in incubation mixtures containing 1 mmol substrate, 0.25 mg protein/mL of HLMs, and 1 mmol NADPH in 0.1 mol/L potassium phosphate buffer (pH 7.4; total volume, 0.4 ml). The reaction was started with the addition of microsomes following a 3-minute preincubation at 37°C, conducted in a shaking water bath at 37°C for 10 minutes, and stopped by the addition of 0.1 mL of ice-cold ethanol. The formaldehyde which then formed was determined by a modification of the method of Jones et al. (21). In detail, after protein precipitation, 0.4 mL of clear supernatant was transferred into a tube containing 80 μL of freshly prepared Nash's reagent [15% (w/v) ammonium acetate, 0.3% (v/v) acetic acid, and 0.2% (v/v) 2,4-pentandione in distilled water]. The tubes were capped and placed in a water bath at 51°C for 12 minutes to allow formation of a formaldehyde derivative (3,5-diacetyl-1,4-dihydrolutidine), which was then separated from interferences and quantified by HPLC as described below.

General Incubation Procedure. A typical incubation mixture consisted of 0.3 mol/L Tris (pH 7.4), 0.5 mmol NADPH, 0.25 mg protein/mL of HLMs and MMDX or [¹⁴C]MMDX (1-50 μmol/L) in a total volume of 0.2 mL. The reactions were started by adding the microsomes following a 3-minute preincubation at 37°C, conducted in a shaking water bath at 37°C under aerobic conditions, and stopped, typically after 10 minutes, by adding 0.2 mL of ice-cold methanol. Control incubations were conducted in the absence of NADPH or using boiled microsomes. After protein precipitation, an aliquot of the supernatant was analyzed by HPLC with fluorescence detection, or combined UV and radiometric detection, or by liquid chromatography with tandem mass spectrometry detection (LC-MS/MS), as described below.

Kinetic Analysis of PNU-159582 Formation. Incubations for analysis of enzyme kinetics were carried out at eight MMDX concentrations ranging from 1 to 50 μmol/L; the incubation protocol was the same as that described above. Enzyme kinetic parameters (K_m and V_max) were determined by fitting the reaction velocity versus substrate concentration data to the single enzyme Michaelis-Menten equation using the SigmaPlot 8.0 software (SPSS, Inc., Chicago, IL).

Correlation Studies. MMDX (20 μmol/L) was incubated with microsomal fractions from 10 individual human livers; the incubation protocol was the same as that described above. The rates of PNU-159682 formation obtained in these experiments were correlated with several known CYP form-selective catalytic activities evaluated in the same microsomal samples (data provided by BD Gentest except those for nifedipine oxidation and erythromycin N-demethylation). Coefficients of determination (r²) and P values were determined by linear regression analysis (GraphPad Instat version 2.02; GraphPad Software, San Diego, CA).

Chemical and Immunochemical Inhibition Studies. Formation of PNU-159682 from 20 μmol/L MMDX by pooled HLMs was evaluated in the absence (i.e., control) and presence of known CYP form-selective chemical inhibitors. The following inhibitors were examined at concentrations previously identified as being appropriate to cause CYP form-selective inhibition in HLMs: 7,8-benzoflavone (1 μmol/L, CYP1A2-selective; refs. 22, 23), sulfaphenazole (20 μmol/L, CYP2C9-selective; ref. 22), quinidine (5 μmol/L, CYP2D6-selective; ref. 22), diethyldithiocarbamate (25 μmol/L; CYP2A6/E1-selective; ref. 24), troleandomycin (100 μmol/L, CYP3A-selective; ref. 22) and ketoconazole (1 μmol/L, CYP3A-selective; ref. 25). In experiments with reversible inhibitors, i.e., 7,8-benzoflavone, quinidine, sulfaphenazole, and ketoconazole, the inhibitor was coincubated with the substrate; the incubation protocol was the same as described above. In experiments with mechanism-based inhibitors, i.e., diethyldithiocarbamate and troleandomycin, the inhibitor was preincubated with liver microsomes and NADPH (0.5 mmol) at 37°C for 15 minutes before adding the substrate and additional 0.5 mmol NADPH. The reactions were then conducted as described above.

Immunochemical inhibition studies were carried out using mouse ascites fluids containing inhibitory MAbs which have been shown to be specific for different human CYP enzymes (15–19). Pooled HLMs (0.25 mg microsomal protein/mL; 20 pmol of total CYP) were preincubated with the designated amount of mouse ascites containing anti-CYP MAb (20-140 μg) at 37°C for 5 minutes in 0.3 mol/L Tris (pH 7.4); the reaction was then initiated by the addition of MMDX (final concentration, 20 μmol/L) and NADPH (final concentration, 0.5 mmol/L) in a total volume of 0.2 mL, and conducted as described above. The highest concentration of each MAb used in these trials (i.e., 7 μg ascites protein/pmol of total CYP) was previously shown to be saturating for an appropriate CYP form-specific reaction in HLMs (15–19). Control incubations were carried out in the absence of MAb.

Incubations of MMDX with microsomes containing cDNA-expressed CYP enzymes were done as described for HLMs, except that the amount of enzyme used was 50 pmol/mL and incubations were terminated after 60 minutes; substrate concentration was 20 μmol/L. All incubations were done in duplicate. Aliquots of the supernatants from each sample were analyzed for PNU-159682 content by HPLC with fluorescence detection.

Radio-High-Performance Liquid Chromatography Analysis. To investigate the overall liver microsomal metabolic profile of MMDX, analysis of incubation mixtures was carried out using an HPLC system consisting of a series 200 LC pump fitted with a solvent degasser (all from Perkin-Elmer Corp., Norwalk, CT), an AS-950 autosampler (Jasco, Inc., Tokyo, Japan), an LC 295 UV-Visible detector (Perkin-Elmer), and a 515TR radioactivity flow detector (Packard Instruments, Meriden, CT) equipped with a 0.5 mL homogeneous flow cell. Chromatographic conditions were as follows: column, Simmetry C₁₈ (4.6 × 150 mm, 3.5 μm; Waters Corp., Milford, MA); mobile phase, 10 mmol KH₂PO₄ (pH 5.5)/methanol/acetonitrile (45:30:25, v/v/v); flow rate, 1 ml/min; injection volume, 20 to 50 μL; column temperature, ambient; detection, UV absorbance at 254 nm and radioactivity using Ultima-Flo M liquid scintillation cocktail (Packard Instruments) at a 3:1 ratio. Data were collected and processed using the Flo-one software (Packard Instruments).

Structural Identification of PNU-159682 by Liquid Chromatography with Tandem Mass Spectrometry Detection. The chromatographic system consisted of a 2690 Alliance separation module including a gradient HPLC pump, an autosampler and a column oven (Waters), a TSP UV 600 variable-wavelength UV detector (ThermoFinnigan, San Jose, CA) and an LCQ classic ion-trap mass spectrometer (ThermoFinnigan). Chromatographic conditions were as follows: column, Xterra MS C₈ (2.1 × 150 mm, 5 μm; Waters); mobile phase, 10 mmol ammonium formate (pH 4.5; solvent A) and acetonitrile (solvent B); elution program, linear gradient from 0% to 70% solvent B from 1 to 10 minutes, followed by isocratic elution with 70% solvent B for an additional 6 minutes; flow rate, 0.3 ml/min; injection volume, 50 μL; column temperature, 40°C; detection, visible absorbance at 490 nm and tandem mass spectrometry. With the above conditions, retention times of MMDX and PNU-159682 were 9.9 and 11.7 minutes, respectively. The mass spectrometer operated under the following conditions: electrospray in positive ion mode; source voltage, 5 kV; capillary voltage, 31 V; capillary temperature, 230°C; tube lens offset, −20 V. Data were collected by full scan mode (scan range of m/z, 175-700) and analyzed by the Xcalibur software package (ThermoFinnigan). Tandem mass spectra were collected using a collision energy of 50% with wideband activation.

Quantitation of PNU-159682. Quantitative evaluation of PNU-159682 formation was carried out using an HPLC system consisting of a model 510 pump, a model 680 gradient controller (all from Waters), a model 7125 injector (Rheodyne), an LS-5 fluorescence spectrometer (Perkin-Elmer) and a C-R3A chromatopac integrator (Shimadzu Corp., Columbia, MD). Chromatographic conditions were as follows: column, Simmetry C₈ (4.6 × 250 mm, 5 μm; Waters); mobile phase, 10 mmol KH₂PO₄ (pH 5.5)/methanol/acetonitrile (45:30:25 v/v/v); flow rate, 1.5 ml/min; injection volume, 20 μL; column temperature, ambient; detection, fluorescence (excitation, 479 nm; emission, 552 nm). Under the above conditions, retention times of MMDX and PNU-159682 were 7.8 and 9.8 minutes, respectively. Quantitation of PNU-159682 was carried out by comparing peak areas to external standard calibration curves constructed daily in the range 0.03 to 1.50 nmol/0.2 mL. The calibration curves were linear over the concentration range studied (r² ≥ 0.98), and the lowest value in the range represented the limit of the quantitation of the assay.

Quantitation of 3,5-Diacetyl-1,4-Dihydrolutidine. Analysis of derivatized formaldehyde (3,5-diacetyl-1,4-dihydrolutidine) was carried out using a Hewlett-Packard series 1050 HPLC system equipped with a degasser, a quaternary pump, an autosampler, and a multiple-wavelength detector (Agilent, formerly Hewlett-Packard GmbH, Germany); chromatographic data were collected and integrated using the Hewlett-Packard ChemStation software (version A.06.03). Chromatographic conditions were as follows: column, Xterra RP₈ (4.6 × 250 mm, 5 μm; Waters); mobile phase, water/acetonitrile (80:20 v/v); flow rate, 1.3 ml/min; injection volume, 100 μL; column temperature, ambient; detection, visible absorbance at 412 nm. Under the above conditions, retention time of 3,5-diacetyl-1,4-dihydrolutidine was 6.3 minutes. Quantitation was conducted by comparing peak areas to external standard calibration curves generated from formaldehyde standards prepared daily in the range 0.4 to 40 nmol/0.4 mL. The calibration curves were linear over the concentration range studied (r² ≥ 0.98), and the lowest value in the range represented the limit of the quantitation of the assay.

CEM, a human lymphoblastic leukemia; Jurkat, a human T cell leukemia; DU145 a human prostatic carcinoma; and HT-29, a human colon carcinoma, were purchased from American Type Culture Collection (Rockville, MD). A2780, a human ovarian carcinoma; L1210, a murine leukemia; and MX-1, a human mammary carcinoma were from the National Cancer Institute (Frederick, MD). EM-2, a human chronic myeloid leukemia, was obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany). CEM cells were grown in Eagle's MEM (Life Technologies, Invitrogen Corporation, Paisley, United Kingdom). Jurkat, EM-2, and A2780 cells were maintained in RPMI 1640 (Life Technologies, Invitrogen Corporation). HT-29 cell line was maintained in Ham's F12 medium (Life Technologies, Invitrogen Corporation). DU145 and L1210 cells were grown in Dulbecco's MEM (BioWittaker, Verviers, Belgium). All culture media were supplemented with 10% (v/v) heat-inactivated FCS and 2 mmol glutamine (both from Life Technologies, Invitrogen Corporation); cultures were maintained at 37°C in a humidified atmosphere containing 5% CO₂. MX-1 human mammary carcinoma was maintained by serial s.c. transplantation of tumor fragments (∼15-20 mg) into the flanks of 4- to 6-week-old female CD-1 athymic nude mice (from Charles River, Calco, Italy).

The cytotoxic effects of doxorubicin, MMDX, and PNU-159682 on adherent tumor cell lines (HT-29, A2780, and DU 145) were evaluated using the sulforhodamine B assay as described by Skehan et al. (26); the effects of the drugs on the growth of nonadherent tumor cell lines (CEM, Jurkat, and EM-2) were evaluated by counting the surviving cells at the end of the treatment period with a ZM Cell Counter (Coulter Electronics, Hialeah, FL). Exponentially growing cells were seeded 24 hours before treatment and exposed to drugs for 1 hour, after which the medium was withdrawn and cells were incubated in a drug-free medium for 72 hours; control cells were not exposed to the drugs. Within each experiment, determinations were done in six times. IC₇₀ values were then calculated from semilogarithmic concentration-response curves by linear interpolation. Data were expressed as mean ± SE of at least three independent experiments.

All procedures involving animals and their care were in conformity with institutional guidelines that comply with national and international laws and policies (European Economic Community Council Directive 86/609, OJ L 358, 1, Dec. 12, 1987; NIH Guide for the Care and Use of Laboratory Animals, NIH Publication no. 85-23, 1985).

Disseminated L1210 Leukemia. Eight-week-old inbred female CD2F1 (BALB/c × DBA/2; from Charles River) were used for evaluation of the therapeutic efficacy of PNU-159682, in comparison with that of MMDX. Disseminated neoplasia was induced by i.v. injection of 10⁵ L1210 cells; 1 day later, the animals were randomly assigned to an experimental group (n = 10) and received a single i.v. injection of MMDX, PNU-159682, or saline (control group). Treatment efficacy was evaluated by comparing the median survival time in the treated and control groups, and expressed as increase in life span as follows: % increase in life span = (100 × median survival time of drug treated mice / median survival time of control mice) − 100. Statistical comparison between the groups was made using the nonparametric Mann-Whitney test (GraphPad Instat version 2.02, GraphPad Software).

Subcutaneous MX-1 Human Mammary Adenocarcinoma Xenografts. Four- to six-week-old female CD-1 athymic nude mice (from Charles River) were used for evaluation of the activity of PNU-159682 against MX-1 human mammary carcinoma xenografts. On day 0, animals (n = 14) were grafted s.c. with MX-1 tumor fragments in the right flank. Eight days later, they were randomly assigned to the drug treatment group or control group (n = 7 mice per group), and treatment was started. PNU-159682 was given i.v. (4 μg/kg) according to a q7dx3 (every 7 days for three doses) schedule; control animals received saline injections. Tumor volume was estimated from measurements done with a caliper using the formula: tumor volume (mm³) = D × d² / 2; where D and d are the longest and the shortest diameters, respectively. For ethical reasons, control animals were sacrificed on day 21 when the mean tumor volume in the group was ∼2,500 mm³; animals receiving drug treatment were monitored up to day 50, at which point they were sacrificed.

Analysis of incubation mixtures containing [¹⁴C]MMDX, NADPH, and native HLMs by reversed-phase HPLC coupled to UV and radiometric detection showed that, apart from [¹⁴C]MMDX, five new UV-absorbing and radioactive peaks, which were undetectable when NADPH was omitted from the mixture or when the incubation was done using heat-inactivated (boiled) microsomes; a typical radiochromatogram is shown in Fig. 1. The main peak eluting at 8 minutes corresponded to unmodified [¹⁴C]MMDX, whereas the peak of the major metabolite (M5) eluted at 11.3 minutes. The retention time of this metabolite was identical to that of authentic synthetic PNU-159682, a liver microsomal metabolite of MMDX we previously identified in the rat, and chemically characterized by mass spectrometry and nuclear magnetic resonance spectroscopy (14). None of the minor metabolite peaks had a retention time identical to that of the available authentic standard of other known metabolites of MMDX, i.e., MMDX N-oxide and 13-dihydro-MMDX. The identity of M5 was further supported by identical retention time, m/z of parent ion and mass spectral fragmentation pattern, as compared with the synthetic standard of PNU-159682 during analysis of reaction mixtures with liquid chromatography with tandem mass spectrometry detection. The molecular weight of the metabolite was 642, 2 a.m.u. lower than that of MMDX. The liquid chromatography with tandem mass spectrometry detection spectrum showed a major ion at m/z 228 and significant fragment ions at m/z 397, 379, 361, and 321 (Fig. 2). Formation of ions at m/z 397, 379, and 361 is in agreement with a fragmentation resulting from cleavage of the glycone moiety with subsequent loss of two molecules of water. The ions at m/z 612 and 230 in MMDX are reduced by 2 a.m.u. to 610 and 228 in PNU-159682, as they result from ions containing the glycone moiety, thus confirming that the loss of 2 a.m.u. occurs in this part of the molecule, as illustrated in Fig. 2.

Preliminary kinetic determinations showed that PNU-159682 formation from MMDX catalyzed by pooled HLMs proceeded linearly with incubation time up to at least 20 minutes and with increasing amounts of microsomal protein up to 0.5 mg/mL. The metabolite formation was then assayed under linear reaction conditions (0.25 mg of microsomal protein/mL; 10 minutes incubation) over a substrate concentration range of 1 to 50 μmol/L. The Eadie-Hofstee plot of data (rate of PNU-159682 formation versus rate of PNU-159682 formation/MMDX concentration) was linear (data not shown), suggesting that a single microsomal enzyme or multiple enzymes with similar apparent K_m values can catalyze conversion of MMDX to PNU-159682. The K_m and V_max values were 16.20 ± 1.18 μmol/L and 2.29 ± 0.07 nmol/minute/mg protein, respectively (mean ± SE of the parameter estimate). Based on these findings, further analyses of the conversion of MMDX to PNU-159682 by HLMs were carried out by incubating MMDX at 20 μmol/L (i.e., at a concentration close to the K_m value for this reaction) with 0.25 mg/mL microsomal protein for 10 minutes at 37°C.

Several approaches were used to identify the enzyme(s) responsible for the biotransformation of MMDX into its major human liver microsomal metabolite. In a first set of experiments, the rates of PNU-159682 formation were evaluated in fully characterized liver microsomes from 10 subjects. Results (Fig. 3A) indicated a 14-fold variability among individuals because rates of metabolite formation ranged between 0.28 ± 0.02 and 3.96 ± 0.11 nmol/minute/mg protein (mean ± SE of three independent experiments), whereas the mean rate (± SE) was 1.64 ± 0.39 nmol/minute/mg protein. The reaction velocity data were then analyzed for correlation with total CYP content and several CYP form-selective activities evaluated in the same microsomal samples. Highly significant correlations were noted between the rate of PNU-159682 formation and three markers of CYP3A4/5 catalytic activity (Fig. 3B , C, and D), that is the rate of testosterone 6β-hydroxylation (r² = 0.9909; P < 0.0001), the rate of nifedipine oxidation (r² = 0.9520; P < 0.0001) and the rate of erythromycin N-demethylation (r² = 0.9542; P < 0.0001). The sample-to-sample variation in the rate of PNU-159682 formation also correlated significantly with total CYP content (r² = 0.8509; P < 0.0001; data not shown). Conversely, no correlation was observed between the rate of PNU-159682 formation and the rates of phenacetin O-deethylation (CYP1A2), coumarin 7-hydroxylation (CYP2A6), S-mephenytoin N-demethylation (CYP2B6), paclitaxel 6α-hydroxylation (CYP2C8), diclofenac 4′-hydroxylation (CYP2C9), S-mephenytoin 4′-hydroxylation (CYP2C19), bufuralol 1′-hydroxylation (CYP2D6), chlorzoxazone 6-hydroxylation (CYP2E1), and lauric acid 12-hydroxylation (CYP4A; r² range from 0.0001 to 0.1234; all P > 0.37).

Subsequent experiments examined the effects of CYP form-selective chemical inhibitors and CYP form-specific inhibitory MAbs on MMDX metabolism to PNU-159682 by HLMs. Among the chemicals tested, only ketoconazole (1 μmol/L) and troleandomycin (100 μmol/L), both selective inhibitors of CYP3A enzymes in HLMs (22, 25), effectively decreased the rate of PNU-159682 formation (81.4% and 94.7% reduction, respectively; Fig. 4A). Sulfaphenazole (20 μmol/L, CYP2C9-selective; ref. 22), quinidine (5 μmol/L, CYP2D6-selective; ref. 22) and diethyldithiocarbamate (25 μmol/L, CYP2A6/2E1-selective; ref. 24) were relatively weak inhibitors of metabolite formation (≤10% reduction of metabolite formation rate), whereas 7,8-benzoflavone (1 μmol/L), a selective inhibitor of CYP1A2 (22, 23) and a stimulator of some reactions supported by CYP3A4 (27, 28), gave a 12.8% increase of the rate of PNU-159682 formation (Fig. 4A); a higher concentration of this flavonoid (30 μmol/L) caused a substantial (2.5-fold) increase of the reaction rate (not shown). Immunoinhibition studies showed a concentration-dependent inhibitory effect of anti-CYP3A4/5 MAb on formation of PNU-159682; at a concentration of 700 μg ascites protein/mL this MAb preparation yielded 83.2% reduction of the rate of metabolite formation (Fig. 4B). By contrast, anti-CYP1A2, anti-CYP2B6, anti-CYP2C8, anti-CYP2C9, anti-CYP2C19, and anti-CYP2D6 MAbs, at their recommended inhibitory concentrations (15–19), had no significant effects on metabolite formation (Fig. 4B).

A last set of reaction phenotyping experiments examined the catalytic competence of microsomal preparations containing individual recombinant human CYP enzymes, in the conversion of MMDX to PNU-159682. Although 20 μmol/L MMDX was metabolized to PNU-159682 by cDNA-expressed CYP3A4 (46 pmol PNU-159682 formed/60 minute/pmol CYP), microsomes containing CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9*1, CYP2C19, CYP2D6*1, CYP2E1, or CYP3A5 did not produce any detectable amount of the metabolite.

The cytotoxicity of PNU-159682 toward six cultured human tumor cell lines of different histotypes were compared with that of MMDX and doxorubicin. Results are summarized in Table 1. IC₇₀ values of PNU-159682 were in the subnanomolar range (0.07-0.58 nmol/L) and noticeably lower than that recorded for both MMDX and doxorubicin; PNU-159682 was 2,360- to 790-fold and 6,420- to 2,100-fold more potent than MMDX and doxorubicin, respectively (Table 1).

An evaluation of the in vivo antitumor activity of PNU-159682, in comparison with that of MMDX, was then conducted in mice bearing disseminated murine L1210 leukemia. Preliminary toxicity dose-finding studies showed that 15 μg/kg i.v. was the single-dose maximum tolerated dose of PNU-159682. The antitumor effect observed after administration of this dose of PNU-159682 (increase in life span = 29%; P < 0.0001 versus control) was comparable (P = 0.31) to that afforded by 90 μg/kg MMDX (increase in life span = 36%; P < 0.0001 versus control), which represents the optimal dose of MMDX using a single-dose schedule i.v. (1, 9).

A further experiment estimated the antitumor activity of PNU-159682 in athymic nude mice bearing MX-1 human mammary carcinoma xenografts. This tumor is refractory to doxorubicin treatment (1, 29) but highly sensitive to MMDX, with most treated animals showing complete response after repeated i.v. or p.o. administration (1). Drug treatment began on day 8 after tumor implantation and consisted of 4 μg/kg PNU-159682 i.v. every 7 days for three times (q7dx3 schedule). As shown in Fig. 5, PNU-159682 was remarkably active against MX-1 xenografts causing tumor regression in all treated animals. Complete tumor response (i.e., no palpable tumor) was observed in four out of seven animals. Nevertheless, as the maximum tolerated dose of PNU-159682 with a q7dx3 schedule was not determined, the complete response rate achievable in this tumor model may be even greater with schedule optimization.

The study described here shows that CYP3A4 is the predominant, if not the only, human CYP enzyme responsible for formation of PNU-159682, i.e., the major liver microsomal metabolite and bioactivation product of MMDX. The conclusion of CYP3A4-selective catalysis is derived from the results of an integrated in vitro approach (30, 31) based on: (a) analysis of the possible correlation between the rates of PNU-159682 formation and several CYP form-selective activities in a panel of HLMs from 10 subjects, (b) evaluation of the effects of CYP form-selective chemical inhibitors and MAbs specific and inhibitory to different human hepatic CYPs, on MMDX conversion to PNU-159682 by pooled HLMs, and (c) evaluation of the intrinsic ability of individual recombinant human CYP enzymes to catalyze formation of the metabolite. Correlation studies showed that the rate of PNU-159682 formation was highly significantly correlated with three different CYP3A-mediated activities (Fig. 3B,, C and D) but did not correlate with marker activities of nine other CYPs in the bank of HLMs. Chemical and immunochemical inhibition of liver microsomal CYP3A resulted in marked (>80%) inhibition of MMDX metabolism to PNU-159682 (Fig. 4). On the other hand, inhibition (by chemicals and/or MAbs) of other CYP enzymes responsible for human drug metabolism (i.e., CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, and CYP2E1) did not result in any significant reduction of the rate of PNU-159682 formation. Further corroborating the role of CYP3A4 in this biotransformation, formation of PNU-159682 was catalyzed only by cDNA-expressed CYP3A4 but not by any of the other nine recombinant CYP enzymes examined, including CYP3A5. Moreover, consistent with the involvement of a single CYP form, conversion of MMDX to PNU-159682 catalyzed by HLMs conformed to single-enzyme Michaelis-Menten kinetics (not shown).

Further in vitro experiments aimed at obtaining information about the cytotoxicity of PNU-159682 toward human tumor cell lines of different histotype confirmed the much higher potency of this compound, compared with MMDX and doxorubicin (Table 1). Moreover, PNU-159682 showed moderate activity against disseminated murine L1210 leukemia in CD2F1 mice (increase in life span = 29%; P < 0.0001 versus control) and remarkable effectiveness toward MX-1 human mammary tumor xenografts in nude mice (Fig. 5); the profile of in vivo antitumor efficacy of PNU-159682 seems therefore to mimic that of its parent compound, MMDX (1, 9). This is not surprising because based on its extremely higher potency compared with MMDX, and requirement of catalysis by CYP3A for its formation, PNU-159682 plays a role in the in vivo antitumor activity and host toxicity of MMDX (9). A role of metabolism in the activity of MMDX in vivo is also suggested by the lack of correlation between drug plasma area under the curve and hematologic and nonhematologic toxicity observed during phase I and II clinical trials (32, 33). Ongoing studies aimed at exploring the molecular mechanism of action of PNU-159682 indicate that it has different effects on cell cycle progression and different DNA-interacting properties, compared with both MMDX and doxorubicin (34). Moreover, further recent data suggest that PNU-159682 retains its activity against tumor cell lines with different mechanisms of resistance to classical anticancer agents, including MDR-1 gene overexpression, reduced topoisomerase II activity, and mutations in the topoisomerase I gene, these latter genetic alterations conferring resistance in vitro to the parent drug, MMDX (35).

The conclusion that CYP3A4 is responsible for the oxidative bioactivation of MMDX in humans is important considering that this CYP contributes, at least partially, to the metabolism of ∼50% of the drugs for which the enzymes responsible for their metabolism have been identified to date, including several anticancer agents (36). CYP3A4 is universally expressed in adult human liver accounting for, on average, 29% of the total hepatic CYP (37) and relatively high levels of the protein (∼50% of hepatic levels) are also present in the small intestine, playing a significant role in the first pass metabolism of various drugs (38). Furthermore, there is high variability in the constitutive levels and activity of hepatic and intestinal CYP3A4 between individuals, which has been reported to range by at least 30-fold (38). It has been recently concluded that genetic factors are of great importance for the interindividual variability in constitutive levels of CYP3A4, accounting for 70% to 90% of the variation, although the underlying mechanisms are still unknown (39). In addition, it is well known that such differences in constitutive CYP3A4 activity can be increased significantly by induction and inhibition of the enzyme by several common drugs, which could therefore cause clinically important variation in MMDX toxicity and efficacy, as previously shown for several other drugs (40). With regard to this, although induction of murine hepatic CYP3A did not cause any significant alteration of the activity of MMDX in vivo, administration of the selective CYP3A inhibitor troleandomycin, markedly decreased the therapeutic effects of the drug in tumor-bearing mice (9). Administration of potent inhibitors of CYP3A4 activity such as azole antifungals (e.g., ketoconazole and itraconazole) and some macrolide antibiotics (e.g., erythromycin and troleandomycin) should therefore be avoided during chemotherapy with MMDX, to reduce the risk of a therapeutic failure due to inefficient oxidative bioactivation of the drug. Monitoring of plasma levels of PNU-159682 during therapy with MMDX should help to individuate the optimum therapeutic concentrations of the metabolite, and allow adjustment of drug doses which will ideally minimize toxicity and maximize successful treatment with MMDX. In vitro drug interaction studies based on evaluation of the inhibitory effects of MMDX and PNU-159682 on hepatic CYP form-specific activities are in progress at our laboratory, to determine whether MMDX may potentially affect the CYP-mediated metabolism of concurrently given drugs.

Some CYP genes including CYP3A4 have been found to be expressed in several human neoplasms, including hepatocellular carcinomas (41), which are poorly responsive to chemotherapy (42). Furthermore, it has been recently shown that microsomes from colorectal cancer samples have the ability to convert paclitaxel to a less cytotoxic metabolite, 3′-p-hydroxypaclitaxel (43); this metabolic reaction is known to be selectively catalyzed by CYP3A4 (44). Besides paclitaxel, CYP3A4 has been shown to be involved in the metabolic inactivation of several other anticancer agents such as docetaxel, Vinca alkaloids, and irinotecan (44–46). It is conceivable that expression of CYP3A4 by tumor cells in vivo may negatively modulate their sensitivity to these agents as it has recently been shown in vitro for paclitaxel, vincristine, and vinblastine, using stably transfected cell lines expressing CYP3A4 (45, 47). By contrast, patients with tumors expressing functional CYP3A4 may benefit from treatment with MMDX due to intratumoral conversion of the drug into a highly potent cytotoxin, PNU-159682.

In conclusion, our findings point to a determinant role of human CYP3A4 in the conversion of MMDX to its highly cytotoxic metabolite, PNU-159682, and suggest that further studies are needed to appreciate the role of intratumoral activation in the patients' response to the drug.

We thank D.J. Waxman, I. Poggesi, and R. Padrini for critically reading the manuscript; H.V. Gelboin for providing us with inhibitory anti-CYP monoclonal antibodies; K.W. Krausz for advice on their use; R.O. Strobl for assistance in preparing the manuscript; and P. Favero for competent technical assistance.