Epo A and B (Epo B, EPO906) were provided by Novartis Pharma [Basel, Switzerland)(purity ≥98% as shown in HPLC), SKF525A by SmithKline Beecham (Brentford, UK), docetaxel and iritotecan (CPT11) by Rhone Poulenc Rorer (Vitry, France), paclitaxel by Bristol Myers Squibb (Wallingford, CT). Sulfaphenazole and anti-CYP3A4 inhibition kit were purchased from Gentest (Woburn, MA), diclofenac, phenacetin, chlorzoxazone, testosterone, midazolam, trichloropropane oxide, erythromycin, ketoconazole, furafylline, ciclosporin A, etoposide, vincristine, vinblastine, verapamil and troleandomycin were obtained from Sigma-Aldrich (St Quentin-Fallavier, France) and ticlopidine from ICN Pharmaceuticals (Orsay, France). HPLC columns were purchased from Interchim (Montluçon, France) and Phenomenex (Torrence, CA).

Liver samples were obtained from adult donors for organ transplantation after approval by the Ethical Committee of the Institut National de la Santé et de la Recherche Médicale. Donors had no drug history and died from traffic accident. No information was available about their smoking habits. Samples were removed within minutes after circulation arrest, immediately frozen into liquid nitrogen and kept at -80°C until use.

Male Sprague Dawley rats (200g) were treated i.p. with phenobarbital in saline (80mg/kg) daily for 3 days. After sacrifice, livers were excised and microsomes prepared as reported earlier [19]. Frozen human livers were thawed into ice-cold isotonic saline, homogenized and processed as rat livers. Microsomes were resuspended into 100 mM sodium phosphate buffer, pH 7.4, 20% glycerol: their cytochrome P450 content was estimated by the procedure of Greim [20] and the protein concentration was measured by a micro BCA kit (Pierce). Monooxygenase activities and the presence of individual CYP proteins were determined in liver microsomes as stated previously [21-24].

Full length cDNAs coding for human CYP1A1, 1A2, 2A6, 2C8, 2C9, 2C18, 2C19, 3A4, 3A5 and 3A7 were inserted in the correct orientation of the expression vector pMT2 as previously reported [22-24]. Ad 293 cells were cotransfected with 50µg of CYP-pMT2 plasmid and 5µg of vector conferring the resistance to geneticin. Full length cDNAs for CYP2B6, 2D6 and 2E1 were cloned in pCIneo (Promega, Madison, WI). Clones growing in presence of G418 and expressing a high level of CYP protein determined by immunoblotting were selected as previously described [22-24]. In case of CYP3A4, 3A5 and 3A7, stable cell lines were further transfected with the full length cDNA encoding human NADPH-cytochrome P450 reductase inserted into pMT2. The addition of multiple copies of the reductase gene markedly increased the production and activity of the reductase protein and consequently stimulated monooxygenase activities supported by CYP3A genes but had no action on the catalytic activity of other CYP proteins. Cells near half-confluence in 75 cm² flasks were incubated for 24h with 50µM Epo in 4 ml Dulbecco-MEM medium. Epo and their metabolites were extracted three times with 5 ml dichloromethane and processed as indicated below.

Rat liver microsomes, corresponding to 1 nmole total P450, were incubated in a final volume of 0.5 ml of 50 mM sodium phosphate buffer pH 7.4, 5 mM MgCl₂, 10% glycerol containing 0.1 mM NADP and 1 mM glucose 6 phosphate. Incubation conditions were selected after optimization for the production of metabolites. Epo A and B dissolved in methanol to give a stock solution at 10 mM, were added to a final concentration of 50 µM (except when otherwise indicated) and the reaction was initiated by the addition of glucose-6-phosphate dehydrogenase. The incubation was carried out at 37°C for 30 minutes and was stopped by the addition of 2 ml dichloromethane. Extraction was repeated twice: organic phases were pooled and evaporated to dryness under nitrogen.

Human liver microsomes (1 mg protein) were similarly processed except that incubations were carried out for 60 minutes. Kinetic parameters were measured in incubations performed following the same protocol with increasing concentrations of Epo ranging from 0 to 100 µM. For inhibition studies, chemicals dissolved in a minimum volume of methanol were added to the incubation mixture in the presence of the NADPH generating system for a 10 min preincubation and the reaction was initiated by the addition of Epo. Immunoinhibition studies with CYP3A4 antibodies were performed following the recommendation of the manufacturer.

Dry residues were dissolved in methanol and analyzed by HPLC. Briefly, aliquots were injected onto a Zorbax C18 column (250x4.6 mm, 5µm) fitted with a guard column. Elution was performed at a flow rate of 1 ml/min with a mixture of 60% methanol, 40% water and was monitored at 254 nm. Under these conditions, retention times for Epo A and B were 14.9 and 18.3 minutes, respectively. For identification of Epo metabolites in LC/MS/MS analysis, samples were injected in the HP1100 model HPLC system (Agilent Technologies, Massy, France) including a binary pump and a by-pass loop autosampler and a microbore C18 Nucleosil column (50 x 1 mm, 5µm) (Interchim, France). The elution was realized at 50µl/min with a gradient of acetonitrile from 20 to 80% in water-0.1%formic acid. The column outlet was connected in line to a UV detector set at 254 nm and the orthogonal electrospray source of a triple quadrupole mass spectrometer (Quattro-LCZ, Micromass, Manchester, UK) operated in precursor ion scan and in product ion scan mode with argon as a collision gas. The collision energy was set at 30V. Positive ion electrospray spectra were recorded over the mass range m/z 80 to 600 in 1 sec. The spray voltage was set at 3500V and the cone voltage of the ion source was kept at 40V.

For analysis of monooxygenase inhibition, pooled human liver microsomes (lot no 0110186, total CYP content of 0.57 nmoles.mg^-1 protein) were purchased from XenoTech (Lenexa, KS). Assays were performed at 37°C in a final volume of 100µl containing 100 mM phosphate buffer (pH 7.4), 5 mM MgCl₂ and 0.2 or 0.5 mg/ml microsomes. Probe substrates dissolved in DMSO (0.2% final concentration) were incubated at the concentrations stated in Table 2. Epo B was added in a concentration range between 1 to 100 µM. Reactions were initiated by the addition of 1 mM NADPH after a 3 min preincubation period and terminated after 10-30 min by addition of 200µl acetonitrile. Samples were evaporated and reconstituted in 100 µl 5/95 (v/v) acetonitrile/water. After addition of internal standard warfarin (measurement in positive ion mode) or 4’-hydroxy-S-mephenytoin (negative ion mode), 10 µl aliquots were analyzed by LC/MS/MS with a Shimadzu liquid chromatograph equipped with a PE Sciex API 3000 (Applied Biosystems, Foster City, CA) mass spectrometer. Chromatographic separation was achieved with a Phenomenex polar RP column (50x2.1mm, 4µm). The mass spectrometer was operated in multiple reaction monitoring (MRM) mode with the Turbo Ion Spray probe at 380°C.

When the incubation mixture of Epo A with liver microsomes isolated from PB-treated rats was subjected to HPLC, two peaks called m1 and m2 eluting at 6.9 and 7.8 min were monitored at 254 nm in addition to the parent molecule which had a retention time of 14.3 min (data not shown). When NADPH was omitted from the incubation, the formation of these products was severely impaired suggesting the participation of CYP proteins. To definitively demonstrate their involvement in the reaction, inhibitors were used to prevent the formation of the derivatives. Trichloropropane oxide (TCPO) selectively inhibits microsomal epoxide hydrolase [25]: when added to the incubation, TCPO failed to reduce the production of Epo A derivatives. Furthermore when Epo A was incubated in the absence of NADPH with microsomal preparations or cytosol, no diol resulting from the hydration of the epoxide by microsomal or cytosolic epoxide hydrolase was formed. This suggested that under these conditions the epoxide was quite stable and that epoxide hydrolase did not play a role in the in vitro biotransformation of Epo A. Conversely, SKF 525A a non-specific inhibitor of CYP, almost completely prevented the formation of the derivatives thus involving CYP proteins.

When incubated with 50µM Epo A, human liver microsomes formed metabolites eluting with the same retention time as the metabolites generated by rat microsomes. The structure of Epo A and its metabolites was elucidated by LC/MS/MS. The mass spectrum of authentic Epo A showed fragment ions at m/z 406, 306, 166 and 113 in addition to the molecular mass ion at (M+H⁺) 494. These fragments resulted from the release of the lateral chain (m/z 166 and 113) or from the breakage of the macrolide ring (m/z 406 and 306). Similarly analyzed, the Epo A derivatives m1 and m2 yielded the same molecular mass of 509 [(M+H⁺) 510]: this increase of 16 of the molecular mass would indicate the addition of an hydroxyl group on the Epo A molecule. This was further evidenced by the loss of two molecules of water (m/z 492 and 474) consistent with the presence of two hydroxyl groups, whereas only one group could be released from the native Epo A molecule (m/z 476). The fragmentation of metabolites m1 and m2 generated fragments at m/z 322 and m/z 422 indicating that the additional hydroxyl group was inserted on the macrolide ring, whereas the lateral chain remained unchanged with fragments at m/z 166. Taken together these data clearly suggest that the hydroxylation of Epo A occurs on the macrolide ring, most likely on carbons 9, 10 or 11.

The rate of oxidative biotransformation of Epo A is related to the CYP content of microsomes: in rat 1 nmole of total P450 hydroxylated roughly 16 nmoles of Epo A in 30 minutes, while 8-9 nmoles of Epo A were transformed by 0.3 nmoles of human P450 in 1 hour. Thus, the turnover number or molecular activity of the monooxygenase system towards Epo A was comparable in rat and human when expressed as nmole metabolite formed by 1 nmole total P450.

When 50 µM Epo B were incubated with human liver microsomes under the same conditions as Epo A, two peaks called m1 and m2 were eluted by HPLC with a retention time of 7.8 and 9.1 min versus 18.3 min for the parent Epo B molecule (data not shown). Examination by LC/MS/MS showed Epo B to produce a molecular mass of 507 [(M+H⁺) 508] with fragments at m/z 420, 320, 166 and 113 (Fig. 1). Its metabolites yielded a (M+H⁺) ion at 524 indicating a monohydroxylation of the parent molecule and the pattern of fragmentation strongly suggested that the hydroxylation site was located on the macrolide ring with fragments at m/z 436, 336, 166 and 113 (Fig. 1). Thus hydroxylation of Epo B catalyzed by human liver microsomes appeared to be quite similar to that of Epo A.

Fig. (1) Mass spectra of Epo B and its metabolites generated by human liver microsomes. Inserts are the proposed fragmentation patterns.

To identify the hydroxylated products of Epo A and B, we also used cultured cell lines expressing individual recombinant CYP proteins. In addition to metabolites m1 and m2 already described, a new metabolite called m3 was isolated from incubations of Epo A and B with CYP2C19 or 2C9 (Fig. 2A). Examination by LC/MS/MS of Epo A m3 and Epo B m3 showed that these metabolites resulted from the monohydroxylation of the parent molecules with an m/z at +16 as for m1 and m2. However the fragmentation pattern appeared different with a fragment at 182 instead of 166 suggesting that the reaction occurred on the lateral chain but not on the thiazole ring with a conserved fragment at m/z 113. The proposed structure of Epo B m3 is shown in Fig. (1).

Fig. (2) HPLC chromatograms of Epo B incubated with human recombinant CYP3A and 2C. Panel A: comparison of chromatograms from incubations of Epo B with CYP2C and CYP3A. Panels B and C: 50µM Epo B was incubated for 24 hrs with cell lines expressing different CYP3A (panel B) or CYP2C (panel C). Culture medium were extracted with dichloromethane, processed as in Material and methods and monitored at 254 nm.

The kinetic parameters of Epo A and B hydroxylation were calculated using rat and human liver microsomes. Km values for Epo A calculated individually in three different human livers ranged from 19 to 38 µM and were 2- to 4-times lower than in microsomes isolated from a pool of PB-treated rats (77µM). The higher P450 content of rat liver microsomes (2.8 nmoles P450/mg) as compared with human liver microsomes (0.38 nmol/mg) accounts for the important difference in the maximal velocity of Epo A transformation existing between rat and human preparations (0.19±0.11 nmol.min^-1.mg^-1 vs 3.8 in rats). Km for Epo B were higher than for Epo A in human samples (74±12µM), but maximal velocities were in the same range (0.24±0.14 nmol.min^-1. mg^-1) suggesting that the affinity of human liver microsomes was slightly lower for Epo B than for Epo A, with identical catalytic efficiencies.

When the oxidative biotransformation of 50 µM Epo A or B was estimated in 16 human liver microsomes previously screened for their monooxygenase activities, a wide interindividual variability in the hydroxylation rates by human liver microsomes was noticed: the formation of metabolites was 31.2±26.5, 84.5±60.6 and 13.3±7.0 pmoles.min^-1.mg^-1 respectively for m1, m2 and m3. The formation of Epo B m1 was highly correlated with the formation of Epo B m2 (r=0.958) but less correlated with the formation of m3 (r=0.862). When compared with the isoform content of microsomes, a positive correlation was noted between the formation of Epo B m2 and the CYP3A content (r=0.86) and its testosterone 6β hydroxylase activity (r=0.74), but not with the CYP2C content (r=0.56). When the biotransformation of taxoids was examined in the same hepatic samples, the correlation existing between the formation of Epo B m2 and CYP 3A dependent activities was high: r=0.873 and 0.733 respectively with the hydroxylation on the lateral chain of docetaxel and paclitaxel, whereas the hydroxylation on the taxane ring of paclitaxel supported by CYP2C8 was only poorly correlated (r=0.45). Thus it appears that the formation of metabolites m1 and m2 by human liver microsomes is predominant and could be attributed to CYP3A isoforms whereas metabolite m3 produced by CYP2C is of minor importance.

To tentatively identify the CYP protein(s) involved in the hydroxylation of Epo A and B, several approaches have been combined. In a first attempt, we used CYP specific inhibitors in three individual human liver microsomal preparations. The formation of individual metabolites m1, m2 and m3 of Epo B was examined and displayed in Table 1: 2µM ketoconazole (inhibitor of CYP3A [26] was a potent inhibitor of the formation of m1 and m2 with 83-100% inhibition, whereas ticlopidine (CYP2C19 inhibitor, [27]) reduced by 50% the production of m3 but has no effect on m1 and m2. Sulfaphenazole (a selective inhibitor of CYP2C9 [26]) and furafylline (CYP1A2 [28]) had no or weak inhibitory effect (<10% inhibition) on either metabolites.

Table 1

Inhibition of the Formation of Oxidative Epo B Metabolites by Various P450 Inhibitors

Then inhibition of Epo B metabolism by immunoglobulins raised against CYP3A4 was tested in human liver microsomes. Increasing amounts of serum containing anti-CYP3A4 immunoglobulins inhibited the hydroxylation of Epo B in a dose-dependent manner to reach more than 90% inhibition (data not shown). This clearly confirmed the role of CYP3A isoforms in the oxidative metabolism of Epo.

To definitively ascertain the involvement of CYP proteins in the hydroxylation of epothilones we used cultured cell lines expressing individual recombinant CYP proteins. The formation of metabolites m1 and m2 was mainly detectable in the culture medium of Ad293 cells stably transfected with CYP3A. As in incubations with human liver microsomes the formation of metabolite m2 exceeded that of metabolite m1. Cells expressing CYP3A4 had the highest transformation rate: when incubated with 50µM Epo B for 24h, the production of m1 and m2 accounted for about 5 nmoles per mg cellular proteins in 24 h. Cell lines expressing CYP3A5 and 3A7 had lower activities (Fig. 2B) as well as those expressing CYP2A6, while CYP2B6 and 1A2 had a marginal activity (Table 2). On the other hand, metabolite m3 was mostly produced by CYP2C isoforms: cell line expressing CYP2C19 had the highest activity followed by CYP2C9 (Fig. 2C). Cell lines expressing CYP2C18 and 2C8 exhibited a marginal activity together with CYP2A6. All other cell lines expressing P450 had no detectable activity (Table 2). These metabolites were isolated and analyzed by LC/MS/MS: their fragmentation profiles were identical with those of metabolites obtained with rat and human liver microsomes and thus confirmed former structures.

Table 2

Formation of Oxidative Epo B Metabolites by Recombinant CYP Proteins Expressed in Ad 293 Cells. Results are Expressed as Pmoles Metabolites Formed in 24hrs Per mg Protein. < : Lower than 10 Pmoles Formed in 24hrs Per mg Protein

Similarly Epo A incubated with Ad293 derived cell lines was hydroxylated mainly by cells expressing 3A or CYP2C19 and 2C9. The production of metabolites m1 and m2 accounted for about 15% of 50µM Epo A added to the culture medium, i.e. 30 nmoles formed in 24h per flask.

All together, these data clearly point out that CYP3A4 is the principal isoform responsible for the hydroxylation of Epo A and B, whereas CYP2C19 has a minor participation.

Epo B was evaluated for its ability to inhibit specific monooxygenase activities catalyzed by human liver microsomes. For that purpose, a pool of microsomes prepared from human livers was used and 8 monooxygenase activities representative of the major CYPs expressed in the liver were assayed in the presence of increasing concentrations of Epo B ranging from 1 to 100 µM. Concentrations eliciting 50% inhibition of the probe substrate activity in the absence of Epo B were calculated and shown in Table 3.

Table 3

Inhibition of Monooxygenase Activities by Epo B. IC₅₀ Values in this Table Indicate the Concentration of Epo B Producing 50%s Inhibition of Probe Substrate Metabolism

Two CYP3A4/5 dependent activities (testosterone 6β-hydroxylation and midazolam 1 hydroxylation) were inhibited by Epo B in a dose-dependent manner as well as the hydroxylation of S-mephenytoin by CYP2C19 (Fig. 3). All three reactions were inhibited by Epo B with an IC₅₀ ≈ 5 µM and the maximal extent of inhibition was higher than 90%. Diclofenac 4’-hydroxylation, a marker for CYP2C9 activity, was moderately inhibited (up to 70%) by Epo B with an IC₅₀ ≈ 25µM, whereas other activities representative of CYP1A2, 2C8, 2D6 and 2E1 showed little or no inhibition by Epo B (IC₅₀ ( 75µM). The inhibition of CYP3A4/5 and CYP2C19 by Epo B was further investigated by the determination of Ki. The 4’-hydroxylation of mephenytoin in pooled liver microsomes was competitively inhibited by Epo B with a K_i of 1.75µM in three separate experiments (data not shown). Similarly, for the 1’-hydroxylation of midazolam a K_i of 1.85µM was calculated.

Fig. (3) Inhibition of monooxygenases in human liver microsomes by Epo B.

To investigate the possible interaction of antineoplastic drugs with the oxidative metabolism of epothilones, we incubated Epo A and B with a pool of human liver microsomes in the presence of a series of drugs used in chemotherapy or with drugs known for their affinity for CYP3A isoforms. Concentrations of Epo A and B chosen for those experiments corresponded to twice the Km for Epo (50 and 120µM respectively for Epo A and B) and for drugs when available from the literature. Results of the competition between Epo B and drugs were shown in Fig. (4). The most potent inhibitor of Epo B hydroxylation was troleandomycin which at a dose of 10µM reduced the formation of hydroxyepothilone B by 95 %. A similar inhibition was observed for the hydroxylation of Epo A. Erythromycin (400µM) and cyclosporin A (25µM) also actively inhibited the formation of the Epo B hydroxyl metabolites. Among antineoplastic drugs, vinblastine (15µM), vincristine (15µM), etoposide (VP16, 100µM) and docetaxel (5µM) exerted a pronounced effect on the hydroxylation of Epo B whereas paclitaxel (30µM) was less active. Irinotecan (CPT11) at a dose of 60 µM had no effect on the oxidative biotransformation of either Epo A or B.

Fig. (4) Effects of various drugs on oxidative Epo B metabolism by human liver microsomes. A pool of human liver microsomes (0.75 nmoles P450) was incubated with 120 µM Epo B in presence of 10µM troleandomycin (TAO), 400µM erythromycin (Eryt), 25µM ciclosporin A (Ci A), 150µM verapamil, 5µM docetaxel, 30µM paclitaxel, 15􀀁M vinblastine, 15µM vincristine, 100µM etoposide (VP16) or 60µM irinotecan. After 60 min the reaction was stopped, Epo B and its metabolite were extracted with dichloromethane and separated by HPLC. Results are calculated as the residual activity and expressed as a percentage of the activity measured in absence of drugs (control).