Experimental Procedures

Materials.

Human liver samples were kindly supplied by Dr. W. Lauchart, Department of Surgery, University of Tuebingen. These were either samples from livers excluded from transplantation for medical reasons or excess normal tissue obtained on partial hepatectomy for tumor metastases. They were cut into pieces of 5 to 10 g and stored at −80°C.

The pamoate salt of racemic ketotifen and (R)-(+)- and(S)-(−)-ketotifen (free bases) were generously donated by Novartis Pharma AG (Basel, Switzerland). β-Glucuronidase/arylsulfatase from Helix pomatia, β-glucuronidase from E. coli, and UDP-glucuronate were purchased from Boehringer (Mannheim, Germany),d-glucuronolactone from Sigma-Aldrich (Deisenhofen, Germany), Isolute SPE (solid-phase extraction) columns filled with 500 mg of the strongly acidic cation exchanger SCX from ict (Bad Homburg, Germany), and HPLC grade acetonitrile from E. Merck (Darmstadt, Germany). Rat β-glucuronidase was prepared from preputial glands of female rats (Tulsiani and Touster, 1978).

Synthesis of Ketotifen Glucuronides.

The procedure of Luo et al. (1992) was modified in the following way. Racemic ketotifen (1 g) was liberated from the pamoate by extracting an alkalinized suspension in water with 10 ml of toluene. In the toluene solution, 1 g of freshly prepared methyl-1-bromo-1-deoxy-2,3,4-tri-O-acetyl-α-d-glucopyranosuronate was dissolved and the solution was added to 20 ml of 0.5 M NaHCO₃. After 2 days of continuous stirring in an atmosphere of nitrogen, the aqueous phase was replaced by a fresh NaHCO₃ solution and 1 g of glucuronic acid derivative was added. This procedure was repeated after another 2 days, and the third aqueous phase was sampled after 2 days. The combined aqueous phases contained conjugate in which the carboxyl group had been liberated by methyl ester hydrolysis. They were extracted three times with 10 ml of diisopropyl ether, adjusted to pH 12.5 to remove the protecting acetyl groups, extracted again with 3 × 10 ml of diisopropyl ether, and extracted once more after adjustment to pH 7 (Luo et al., 1992). The aqueous solution was passed through a column with 5 ml of C₁₈-silica gel (Polygoprep 60–50 C₁₈; Macherey-Nagel, Düren, Germany), which was washed with 8 ml of water/methanol (9:1, v/v) and eluted with 10 ml of methanol/water (9:1). The residue of the eluate was purified by thin-layer chromatography on two 20 × 20 cm sheets coated with silica gel (Alugram Sil G/UV₂₅₄; Macherey-Nagel) in 1-butanol/acetic acid/water (4:1:1, v/v/v). The UV-absorbing band at R_F 0.35 was extracted with methanol and the extract evaporated under a stream of nitrogen. Glucuronides of (R)- and (S)-ketotifen were synthesized in the same way on a smaller scale, starting from 20 mg of drug and 0.1 g of glucuronic acid derivative in 1 ml of toluene stirred with 2 ml of 0.5 M NaHCO₃. The molar quantities and synthesis yields were determined fromE₃₀₀ − E₃₃₀ of an aliquot in methanol, assuming that the molar absorption difference was the same as that of ketotifen (10,500 cm⁻¹M⁻¹). From racemic ketotifen, 0.21% was recovered as N-glucuronides.

Microsome Preparation.

Liver homogenates were prepared with four volumes of buffer (250 mM sucrose, 20 mM Tris-HCl, 5 mM EDTA, adjusted to pH 7.4 at 37°C) and microsomes were obtained by fractionated centrifugation in the last step at 85,000g for 1 h. They were washed by resuspension in the same buffer followed by centrifugation. Pellets were suspended in buffer (0.6 ml per gram of liver, resulting in about 20 mg protein/ml) and aliquots of 0.2 ml were frozen in liquid nitrogen and stored at −80°C. Protein was measured according to Lowry et al. (1951) with BSA as standard.

Binding Experiments.

Binding of (R)- and (S)-ketotifen to microsomal protein was determined by equilibrium dialysis (Brinkschulte and Breyer-Pfaff, 1979) at 37°C in 57 mM Tris-HCl pH 8.0 containing 5 mM MgCl₂. The enantiomers (50 μM) were added to one 1-ml half-cell, and the concentrations in both cells were measured by HPLC after 2, 4, and 6 h. In experiments without microsomes, equal concentrations in the two cells were achieved after 4 and 6 h. Therefore, dialysis time was fixed at 5 h. In experiments with variable ketotifen concentrations (2–100 μM), the concentration of microsomal protein was 0.5 mg/ml, and when the microsome quantity was varied between 0.2 and 1 mg of protein/ml, the ketotifen concentration was constant at 50 μM. Ketotifen was added to the microsome suspension, and after dialysis was measured in both half-cells by HPLC (see below). Samples were adjusted to 0.2 N HClO₄, after those with ketotifen added to 50 or 100 μM had been diluted with water 3- and 5-fold, respectively.

Glucuronidation Assay.

Standard incubation mixtures of 1 ml containing 57 mM Tris-HCl pH 8.0, 5 mM MgCl₂, 2 mM UDP-glucuronate, 0.02% Triton X-100, 0.5 to 200 μM (R)- or (S)-ketotifen, and 0.5 mg/ml microsomal protein were shaken for 25 min at 37°C. The reaction was stopped by three extractions with 1 ml oftert.-butyl methyl ether and one extraction with 1 ml ofn-hexane, care being taken to remove the protein interphase with the organic layer. A weighed aliquot of the aqueous phase was mixed with 10% of its volume of 2 N perchloric acid, centrifuged at 10,000 rpm after cooling on ice, and 0.5 ml was injected for HPLC. WhenN-glucuronide mixtures were incubated with microsomes in the absence of UDP-glucuronate, extracted, and analyzed by HPLC after HClO₄ addition, mean recoveries were 101 and 97%, respectively, in duplicate experiments with the glucuronides of(R)- and (S)-ketotifen. Therefore, the quantities measured were not corrected for recoveries. In kinetic measurements, all incubations were carried out in duplicate and means were used for calculations. The coefficients of variation within series with 10 to 11 different substrate concentrations varied between 4 and 11% (mean 7%).

HPLC Analyses.

For N-glucuronide analyses, samples of 0.5 ml were applied to a 20 × 4.6 mm clean-up column with C₁₈-silica gel (Grom-Sil 120 ODS-4 HE, 11 μm; Grom, Herrenberg, Germany) by pumping water (1 ml/min) for 2 min. Samples were transferred to the analytical column (Prodigy 5 μm ODS (3) 100 Å, 250 × 4.6 mm; Phenomenex, Hösbach, Germany) by running the eluent (10 mM sodium phosphate buffer pH 6/acetonitrile, 78:22, v/v, 1 ml/min) for 3 min in the reverse direction. The clean-up column was conditioned with methanol for 0.5 min and with water for 2 min. The eluate was monitored at 300 nm and data were registered by the MT2 integration program (Kontron Instruments, München, Germany). Quantities of individual glucuronides were calculated from peak areas relative to external standards with known total concentrations. It was assumed that the contribution of each glucuronide to total peak area corresponded to its relative quantity.

For recording spectra during the HPLC run, the pump was stopped close to the peak maximum and the absorption was measured between 240 and 340 nm with a UVIS-205 detector (Linear Instruments, Reno, NV).

For ¹H NMR and mass spectrometric investigations, about 0.4 μmol each of the two isomeric(S)-ketotifen glucuronides were obtained from the glucuronide mixture synthesized from racemic ketotifen. Quantities of about 0.1 μmol were injected for HPLC, and eluates corresponding to the first and fourth peaks were sampled separately. Eluates from 18 consecutive injections were combined and concentrated on C₁₈-silica gel (cartridges Bond Elut C18/OH; ict, Frankfurt, Germany), from which the individual glucuronides were eluted with methanol.

Free ketotifen was analyzed in a system consisting of a 6 × 5 mm clean-up column (Polygoprep 60–50 C18; Macherey-Nagel) and a 250 × 4.6 mm analytical column filled with C₁₈-silica gel (Nucleosil 5 C₁₈, Macherey-Nagel). Samples of 0.2 ml were applied to the clean-up column with 30 mM perchloric acid adjusted to pH 2.5 with NaOH, flow 1.5 ml/min, for 2 min. The eluent (10 mM perchloric acid buffered to pH 2.5/acetonitrile 70:30, v/v, 1.2 ml/min) was run in reverse direction for 1 min. The clean-up column was conditioned with 30 mM perchloric acid, pH 2.5, for 2.5 min. Ketotifen (R_T 14 min) was detected at 300 nm and its quantity was derived from the peak height relative to those of external standards.

Enzymatic Hydrolysis of Ketotifen N-Glucuronides.

Incubations with β-glucuronidase/arylsulfatase from Helix pomatia or β-glucuronidase from rat were carried out in 0.1 N sodium acetate/acetic acid pH 5, those with β-glucuronidase fromE. coli in 75 mM potassium phosphate buffer pH 6.8. Rat andE. coli glucuronidase activities were standardized with phenolphthalein glucuronide (Stahl and Fishman, 1984) and specific activities were found to be 1.4 and 0.16 U/mg protein, respectively, at 37°C (producer information on the E. coli enzyme: 20 U/mg with 4-nitrophenyl-β-d-glucuronide as substrate). Quantities used for measuring the rate of ketotifenN-glucuronide hydrolysis were 0.1 to 1 mU/ml of the E. coli enzyme and 5 to 10 mU/ml of that from rat. Samples were drawn after 1 to 8 h at 37°C and analyzed by HPLC for unhydrolyzed glucuronides.

For confirmation of structure, an excess of the Helix pomatia enzyme (0.1 ml) was incubated with 3.5 nmol(S)- or (R,S)-ketotifen glucuronides for 12 h at 56°C and the liberated ketotifen was extracted from the alkalinized incubate and measured by HPLC.

Ketotifen N-Glucuronidation In Vivo.

Five female and five male healthy volunteers took 1 capsule of ketotifen (Zaditen; Sandoz) containing 1.38 mg of(R,S)-ketotifen hydrogen fumarate (corresponding to 1 mg free base) with 50 ml of water at bedtime. They collected urine in three 8-h fractions that were analyzed for N-glucuronides by a modification of the procedure of Fischer and Breyer-Pfaff (1995). Urine (3–5 ml) was diluted with three volumes of 10 mM sodium phosphate pH 4.0, the pH was adjusted to 4.0 and the solution was passed within 13 min through an SCX column, which was subsequently washed with 2 ml 10 mM sodium phosphate pH 4.0, 2 ml of water, and 4 ml of methanol, and slowly eluted with 6 ml of the following mixture: methanol/0.5 M ammonium acetate adjusted to pH 8.0 with ammonia and containing 0.15% triethylamine (4:1, v/v). The eluate was concentrated to about 60% under a stream of nitrogen and evaporated to dryness under reduced pressure at 35°C. The residue was dissolved in 1 ml of 0.2 N HClO₄, of which 0.5 ml was analyzed by HPLC. Recovery experiments were carried out with synthetic(R,S)-ketotifen N-glucuronide at a concentration of 1.5 nmol/ml urine. With fresh columns, the N-glucuronide recovery was 88 ± 5%. Before use, columns had to be conditioned by washing with 3 ml each of methanol, water, and 10 mM sodium phosphate pH 4.0.

Mass Spectrometry.

Mass spectra of (R)- and (S)-ketotifen and of the two glucuronides synthesized from (S)-ketotifen were recorded with a TSQ 700 triple quadrupole mass spectrometer (Finnigan MAT, San Jose, CA) and Finnigan acquisition software in the electrospray ionization (ESI) and collision-induced dissociation (CID) modes. For ESI and CID mass spectra the samples were dissolved in methanol/water (9:1) to concentrations of 10 to 20 ng/μl. These solutions were infused via a syringe pump at a flow rate of 1.5 μl/min into the ion source. The positive and negative ion electrospray needle voltages were +4500 and −3500 V, respectively. The temperature of the heated transfer capillary was set to 120°C. Sheath gas was nitrogen. Spectra were acquired over the mass range 100 to 700 amu in 2 s. The acquisition time was 1 min and the recorded spectra were averaged.

In the CID mass spectrometry mode, argon was used as collision gas. The collision cell pressure was 1.9 mtorr and the collision offset voltage was −24 eV for the N-glucuronides and −28 eV for ketotifen. The scan range was between 20 and 495 amu, the scan time 1.5 s, and the acquisition time 2 min. The recorded spectra were averaged.

NMR Spectroscopy.

The 600-MHz ¹H NMR spectra were recorded on a Bruker DRX 600 at 303°K using Bruker standard software and pulse programs. The ¹H data were referenced to a trace of tetramethylsilane (0.00 ppm). Samples were dissolved in dimethyl sulfoxide-d₆ to concentrations of 4 and 2 mg/0.5 ml for (R)- and (S)-ketotifen and about 0.05 mg/0.5 ml for the N-glucuronides. Hartman-Hahn correlation spectroscopy (HH COSY) and Hartman-Hahn rotating-frame Overhauser effect spectroscopy (HH ROESY) experiments were included as two-dimensional measurements. Data acquisition was performed with sweep widths of 4167 to 6250 (COSY) or 6250 to 11363 Hz (ROESY) and relaxation delays of 5 to 6 or 7.5 to 11 s, respectively.

Calculations.

Kinetic parameters were calculated according to the Michaelis-Menten equation for one or two enzymes by nonlinear least-squares regression analysis (Fig. P; Biosoft, Cambridge, UK). Intrinsic clearance (Cl_int) was calculated asV_max/K_M. Variations given are S.D.

Results

Synthesis and Characterization of KetotifenN-Glucuronides.

Under the conditions used for other antihistamines (Luo et al., 1992), ketotifen could be derivatized at the tertiary amino group with the production of quaternary ammonium salts, although at a poor yield. Although two diastereomeric glucuronides had been expected to be formed from the racemic drug, four reaction products were separated under optimal HPLC conditions. Records of their UV spectra run during HPLC were identical with each other and with that of ketotifen exhibiting maxima at 300 nm. Incubation with β-glucuronidase/arylsulfatase led to the disappearance of all four peaks and the formation of ketotifen, indicating that the peaks contained glucuronides. Syntheses starting from (R)- or (S)-ketotifen resulted in the glucuronides with R_T 13.5 and 13.85 min [designated (R)-Gluc_A and(R)-Gluc_B] or those withR_T 12.4 and 14.5 min [(S)-Gluc_A and(S)-Gluc_B], respectively. Syntheses from racemic ketotifen yielded the N-glucuronides of the two enantiomers in nearly equal quantities. On the other hand, the ratios between the two isomers derived from either (R)- or(S)-ketotifen were variable, but never exceeded 2 whether the racemic drug or one of the enantiomers was derivatized.

To elucidate the difference between(S)-Gluc_A and(S)-Gluc_B, quantities sufficient for instrumental analysis were collected from HPLC eluates. Their¹H NMR and mass spectra were compared with those of (R)- and (S)-ketotifen, the latter two giving identical spectra throughout. In the ESI positive ion mode, the mass spectrum of ketotifen shows the expected [M+H]⁺ion at m/z 310 as the base peak and in the negative ion mode at m/z 308 as [M−H]⁻ (in accordance with the sum formula C₁₉H₁₉NOS). The CID mass spectrum of the precursor ion m/z 310 shows besides the base peak at m/z 96 [C₆H₁₀N]⁺only two noticeable ions of low abundance, m/z213 [M+H−97(C₆H₁₁N)]⁺and m/z 82 [C₅H₈N]⁺. In the ESI mode, the two N-glucuronides exhibit nearly identical mass spectra with molecular ions [M+H]⁺ at m/z 486, sodium adduct ions [M+Na]⁺ atm/z 508, and [M−H]⁻ions at m/z 484. These are compatible with the sum formula C₂₅H₂₇NO₇S. The CID spectra of the two N-glucuronides show identical fragmentation patterns. The precursor ions m/z486 eliminate 176 amu [glucuronic acid−H₂O] with formation of the base peak at m/z 310 [ketotifen+H]⁺. Additional fragment ions of low abundance are m/z 159 [glucuronic acid−2H₂O+H]⁺,m/z 131 [159−CO]⁺,m/z 113 [159−HCOOH]⁺, and m/z 96 [C₆H₁₀N]⁺.

¹H NMR data are presented in Table1. The proton signals of the piperidylidene ring of ketotifen form two groups because the protons on the side directed toward the benzene ring (H-13 and H-14, see Fig. 1) are more shielded and may underlie the magnetic field produced by the ring current. The signals of axial (ax) and equatorial (eq) protons of both groups exhibit identical coupling patterns and can be assigned on the basis of their coupling constants and cross signals in COSY. The eq protons at C-11 and C-14 are influenced by the magnetic anisotropy of the exocyclic double bond and give signals upfield to those of the axial protons. ROESY experiments demonstrate the close steric proximity of H-3 on the thiophene ring to the two hydrogens at C-11 by nearly equally intensive cross-peaks. On the other hand, those between H-5 (and H-6) on the benzene ring and H-14ax and H-14eq are weak (or very weak) and hardly exceed the H-3 to H-12eq signal in intensity. This leads to the conclusion that the axes C-11 to C-12 and C-14 to C-13 are not aligned parallel to the theoretical axis C-4a to C-10, but rather at a right angle. The piperidylidene ring is apparently present in a rigid chair form.

Very intense ROESY cross-peaks indicate a close proximity of H-8 and H-9a, which should be in the plane of the benzene ring, whereas H-9b is on the convex side of the tricyclic system. The keto group is oriented in the thiophene ring plane.

(S)-Gluc_A exhibits small differences from ketotifen with regard to the signals of the aromatic protons and those of H-9a and H-9b. In the piperidylidene ring, the methylene groups in α-position to the quaternary nitrogen atom (H-12 and H-13) are deshielded by 1.17 to 1.55 ppm and those in β-position (H-11 and H-14) by 0.20 to 0.71 ppm. Geminal (²J) and vicinal ax-ax (³J) coupling constants of the methylene groups exceed those of ketotifen, whereas the ax-eq and eq-eq (³J) coupling constants are nearly the same. A symmetrical chair or boat form can be excluded because according to the equal intensity of the ROESY cross-peaks between H-3 and H-11eq on the one hand and H-5 and H-14ax on the other, these distances should be similar (Fig. 2A). A strong cross-peak is also registered between H-1′ and H-14ax, indicating a pseudoaxial or isoclinal position of the glucuronosyl C-1′ on the quaternary ring nitrogen. A close proximity between H-1′ and H-13eq is indicated by a weak cross-peak; therefore, the plane of the glucuronosyl ring should be oriented vertically to the benzene ring. The most probable conformation of the piperidylidene ring is a twist form analogous to that ofcis-1-tert.-butyl-4-phthalimidocyclohexane (Kellie and Riddell, 1974). In the present case, a twist form may be further favored by the presence of a sp²-hybridized carbon atom in the piperidylidene ring.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Structural formulae showing the conformations of the (S)-ketotifen N-glucuronides (S)-Gluc _A (A) and (S)-Gluc_B (B).

Close steric proximities giving rise to cross signals in ROESY experiments are indicated by dotted lines.

In (S)-Gluc_B the deshielding effect of the quaternary nitrogen amounts to 1.05 to 1.65 ppm in H-12 and H-13 and to 0.28 to 0.47 ppm in H-11 and H-14. An unusual finding is the identity of the shifts of H-13ax and H-13eq, probably caused by the ring torsion that removes H-13ax from the extra magnetic field produced by the ring current of the benzene ring. In comparison with(S)-Gluc_A, H-1′ and H-5′ show upfield shifts of 0.47 and 0.24 ppm, respectively (Table 1), and indicate the influence of anisotropic magnetic fields in their neighborhood. As in ketotifen, the ROESY experiment indicates nearly identical H-3 to H-11ax and H-3 to H-11eq distances. In addition, the small internuclear distance of H-1′ and H-13(ax or eq) becomes apparent from a strong cross-peak, and a weaker peak points to the proximity of H-1′ and H-14eq (Fig. 3). Therefore, C-1′ should be attached axially to the quaternary nitrogen and the piperidylidene ring should be present in a distorted boat form. This is in accordance with the ³J coupling constants that are increased relative to those of ketotifen in ax-eq and eq-eq interactions. The H-13ax to H-14ax constant is smaller, but ³J for H-11ax to H12ax is larger and the geminal (²J) values for H-11 and H-14 are distinctly larger. The glucuronosyl residue is assumed to be placed parallel to the plane of the seven-membered ring on the convex side of the tricyclus (Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Two-dimensional NMR ROESY spectrum of the ketotifen N-glucuronide (S)-Gluc _B .

Arrows point to cross-peaks providing decisive structural information.

Chemical Stability and Enzymatic Hydrolysis of KetotifenN-Glucuronides.

The synthetic glucuronides (S)-Gluc_Aand (S)-Gluc_B were not hydrolyzed on heating to 56°C in 1 N sulfuric acid for 30 min nor did an interconversion or a conversion to (R)-ketotifen glucuronides take place. Also in neutral medium, the compounds were sterically stable during 8 h at 37°C, and samples illuminated by a neon tube for 24 h were not converted to isomers. No racemization of the ketotifen moiety was observed on storage of the glucuronides at 4°C in the dark for several months.

Partial hydrolysis by limiting quantities of β-glucuronidase from E. coli or rat usually was not linear with time. In most cases, the rate was maximal within the first 1 or 2 h with progressive slowing in the following 3 or 6 h; on the other hand, hydrolysis rates increased linearly with enzyme concentrations. Differential hydrolysis took place when the two(S)-ketotifen glucuronides were incubated with E. coli β-glucuronidase. As an example, Fig.4B demonstrates the 26% hydrolysis of(S)-Gluc_A within 1 h that is increased to 42% up to 4 h, whereas the small fraction of(S)-Gluc_B hydrolyzed became measurable only after 4 h (7%). On the other hand, the two(R)-ketotifen glucuronides were split at comparable rates (Fig. 4A). β-Glucuronidase from rat was about 40-fold less active than the bacterial enzyme in N-glucuronide hydrolysis relative to its activity toward phenolphthalein. Incubation of 5.8 μM(S)-ketotifen glucuronides with 5.1 mU/ml for 2 h led to 25 and 18% hydrolysis, respectively, of(S)-Gluc_A and(S)-Gluc_B, and little more reaction up to 8 h. Under the same conditions,(R)-Gluc_A and(R)-Gluc_B were hydrolyzed by 9 and 21%, respectively, within 2 h.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Time course of the hydrolysis of (R)- and (S)-ketotifen N-glucuronides with β-glucuronidase from E. coli.

A, (R)-ketotifen N-glucuronide (6.6 μM) was incubated with 0.16 mU/ml of the enzyme (assayed with phenolphthalein glucuronide); B, (S)-ketotifenN-glucuronide (25 μM) was incubated with 0.32 mU/ml enzyme.

Ketotifen N-Glucuronidation in Liver Microsomes.

Because only substrate not bound nonspecifically to a microsomal suspension is available for metabolism (Obach, 1997), the free fractions of the ketotifen enantiomers were determined under the conditions of microsomal incubations, but with omission of UDP-glucuronate and Triton X-100. A low degree of binding that was independent of the protein concentration in the range 0.2 to 1 mg/ml and of the drug concentration in the range 2 to 100 μM was observed. The main free fraction of (R)-ketotifen was 91 ± 5% and that of the (S)-enantiomer 94 ± 6% (n = 9). These fractions were used for correcting the substrate concentrations added to microsomal incubates.

For in vitro glucuronidation experiments, optimal conditions were determined as shown in Table 2. Glucuronidation rates increased with pH, but values above 8.0 were not tested. Small enhancements were observed on increasing the UDP-glucuronic acid concentration above 2 mM. Addition of 0.01 or 0.02% Triton X-100 stimulated the reaction distinctly, but at 0.03% the rates declined. Reaction rates increased linearly with protein concentrations between 0.25 and 0.75 mg/ml, and they were constant with time up to 40 min.

On incubation of about 5 or 100 μM racemic ketotifen, all four isomeric N-glucuronides were produced (Fig.5). Their identities were confirmed by recording their UV spectra and by enzymatic hydrolysis, which resulted in the disappearance of the glucuronides and a recovery of 78% as ketotifen. No peaks interfering with N-glucuronide analysis were detected in control samples incubated without UDP-glucuronate.(S)-Ketotifen glucuronides represented about two-thirds of the total quantity with (S)-Gluc_Bexceeding (S)-Gluc_A 2- to 3-fold.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

HPLC record of the N-glucuronide mixture produced on incubation of 4.8 μM racemic ketotifen with human liver microsomes and UDP-glucuronic acid in the presence of Triton X-100.

1: (S)-Gluc_A, 2:(R)-Gluc_A, 3:(R)-Gluc_B, 4:(S)-Gluc_B. TotalN-glucuronide quantity was 1 μmol/ml and the ratio of the individual conjugates 20:9:26:45.

For kinetic measurements, (R)- and (S)-ketotifen were incubated separately in the concentration range 0.5 to 200 μM. No higher concentrations were used because in preliminary experiments substrate inhibition had become apparent. N-Glucuronide quantities produced from 400 μM (R)- or(S)-ketotifen amounted to about two-thirds of those measured with 200 μM substrate. In each sample obtained from one of the enantiomers, two isomeric N-glucuronides were measured with mean(R)-Gluc_B/(R)-Gluc_Aratios around 2 and(S)-Gluc_B/(S)-Gluc_Aratios that increased with increasing substrate concentrations from 1.5–2 to 2.3–3.5. Kinetic calculations were based on the sums of the two isomers.

The concentration dependence of production of N-glucuronides from (S)-ketotifen in microsomes from four livers could be depicted by Michaelis-Menten kinetics with two enzymes, one of them exhibiting a very high affinity (K_M1 1.3 μM, Table 3). This can be visualized by Eadie-Hofstee plots (Fig. 6). Omission of Triton X-100 from the incubation medium reducedK_M1, V_max1, andV_max2 values in a comparative experiment with HL 23 microsomes (Table 3), but kinetics remained biphasic (Fig.6). Nonlinear regression coefficients were higher in all cases when parameters were fitted to two-enzyme kinetics (Table 3) than to one-enzyme kinetics. Apparent intrinsic clearances (V_max/K_M) of the high-affinity enzyme exceeded those of the low-affinity component, on average 9-fold, in experiments with detergent.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Eadie-Hofstee plots of the kinetics of N-glucuronide formation from (S)-ketotifen with microsomes from three human liver samples.

HL 23 microsomes were incubated in the presence and absence of Triton X-100, the others in its presence only. The lines were calculated from the parameters for two-enzyme Michaelis-Menten kinetics listed in Table3.

(R)-Ketotifen N-glucuronidation kinetics were more complex, because in incubations without Triton X-100 a contribution of a high-affinity enzyme became apparent that in the presence of Triton was not discernible in two of three microsomal preparations (Table 4, Fig.7). Parameters for two-enzyme kinetics have to be regarded as estimates, because the differences betweenK_M1 and K_M2were too small for exact evaluations of the individual components. As with (S)-ketotifen, the contribution of the high-affinity enzyme to total intrinsic clearance in the absence of Triton X-100 was higher than that of the low-affinity enzyme. In experiments with Triton, one-enzyme kinetics were applicable, resulting in a meanK_M value of 15 μM and, thus, intermediate between those of the two components without detergent, and a very similar intrinsic clearance in spite of a higherV_max value (Table 4). When intrinsic clearances in the presence of Triton were compared, their sum with(S)-ketotifen as the substrate was on average about 50% higher than the value for (R)-ketotifen.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Eadie-Hofstee plots of the kinetics of N-glucuronide formation from (R)-ketotifen with microsomes from two human liver samples in the presence and absence of Triton X-100.

Solid lines were calculated from the parameters for two-enzyme Michaelis-Menten kinetics and dashed lines with those for one-enzyme kinetics listed in Table 4. In the case of HL 18 with detergent, one-enzyme kinetics only were applicable and are depicted by a solid line.

Ketotifen N-Glucuronidation in vivo.

After ingestion of 1 mg of racemic ketotifen, 10 healthy volunteers excreted in urine all four N-glucuronide isomers, but(S)-Gluc_B exceeded the others by far in quantity (Fig. 8, Table5). Within 24 h, urinary glucuronides corresponded to 17.3% of the dose on average, with about 7.6, 6.4, and 3.3% being excreted in the 0- to 8-, 8- to 16-, and 16- to 24-h intervals. This indicates that excretion was not complete after 24 h. The same conclusion can be drawn from the observation that subject WG excreted a higher percentage (29.9%) under continuous dosing with 1 mg ketotifen/day than in the acute experiment (22.9%, Table 5). No significant correlation existed between the percentage of the dose excreted as N-glucuronides and the urine volume.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

HPLC record of ketotifen N-glucuronides in urine of volunteer BP collected 8 to 16 h after ingestion of 1 mg of racemic ketotifen.

Of the extract from 4 ml of urine, an aliquot of 50% was injected. Quantities found were (S)-Gluc_A (1) 0.11 nmol, (R)-Gluc_A +(R)-Gluc_B (2 + 3) 0.07 nmol, and(S)-Gluc_B (4) 0.56 nmol.

To probe chiral stability of ketotifen and its glucuronides, two subjects (BP and MU) took 1 mg (R)-ketotifen as the free base. Their 24-h excretion of (R)-ketotifen glucuronides amounted to 6 and 0.9% of the dose, and in addition, 3.3 and 0.8%, respectively, were found as (S)-ketotifen glucuronides.

Thus, quaternary ammonium glucuronides excreted in urine were mainly derived from (S)-ketotifen, of which a mean fraction of 26% was found as (S)-Gluc_B and 3% as (S)-Gluc_A, whereas on average less than 6% of the (R)-ketotifen dose was glucuronidated in experiments with 1 mg of the racemic drug as well as with 1 mg of the(R)-enantiomer alone.

Discussion

Diastereomeric quaternary ammonium glucuronides were obtained by chemical synthesis or in vitro biosynthesis in human liver microsomes from racemic ketotifen. Unexpectedly, each one of the diastereomers occurred in two conformations that, according to experiments with the (S)-ketotifen derivatives, were not interconverted at temperatures up to 37°C in neutral media nor at 56°C in acid. The same four isomeric N-glucuronides were excreted by humans after ingestion of the racemic drug. Detailed¹H NMR analysis led to the conclusion that differences in the conformation of the piperidylidene ring are the basis for different arrangements of the glucuronosyl group relative to the tricyclic system. As a six-membered saturated ring, the piperidylidene ring may adopt chair as well as nonchair conformations depending on the substituents (Kellie and Riddell, 1974). In ketotifen,¹H NMR data were in favor of a rigid chair conformation, whereas in the N-glucuronide(S)-Gluc_B the data pointed to a distorted boat form, and in (S)-Gluc_Ato a twist form. As a consequence, the glucuronosyl moiety can be assumed to be directed parallel to the tricyclus in(S)-Gluc_B (Fig. 2B), and this may be the reason for the relatively lipophilic character manifested by a longer retention time in reversed phase HPLC than for(S)-Gluc_A. In the latter, the hydrophilic glucuronosyl group is less protected (Fig. 2A); this may lead to faster elution (Fig. 5). Because of the poorer resolution between the (R)-ketotifen N-glucuronide isomers, separate investigation was not possible. An alternative possibility for the occurrence of isomers is the chiral character of the quaternary nitrogen in the N-glucuronides. However, the NMR data allowed us to exclude that (S)-Gluc_Aand (S)-Gluc_B are diastereomers, because in this case one would expect nearly identical chemical shifts and coupling constants for the piperidylidene ring protons.

An isomerism as it was revealed now has to our knowledge not been described for other quaternary N-glucuronides. Two isomers were separated on HPLC of the tertiary 10-N-glucuronide of the tricyclic neuroleptic olanzapine (Kassahun et al., 1997, 1998). The nature of the isomerism has apparently not been elucidated.

The ketotifen N-glucuronides proved resistant toward acid hydrolysis in accordance with the behavior of other drug-derived quaternary ammonium glucuronides (Breyer-Pfaff et al., 1990; Hawes, 1998). The rate of their enzymatic hydrolysis relative to that of phenolphthalein glucuronide was highest with β-glucuronidase fromE. coli. This has been observed with the glucuronides of amitriptyline and diphenhydramine (Fischer and Breyer-Pfaff, 1995) and confirmed for those of chlorpromazine, doxepin, and cyclizine (Hawes, 1998). The bacterial enzyme also proved considerably more active than one from rat toward the ketotifen N-glucuronides, although differences were apparent among the isomers. The hydrolysis of(S)-Gluc_B proceeded much slower than that of (S)-Gluc_A (Fig. 4B), possibly due to a poorer accessibility of the glycosidic linkage. Enzymes produced by E. coli and other bacteria can be assumed to hydrolyze N-glucuronides synthesized in liver and secreted via bile into the intestine, thus leading to enterohepatic drug cycling. Whether this plays a role for ketotifen kinetics in vivo is not clear, but several observations argue in favor of such a process. First, ketotifen plasma concentration profiles sometimes exhibited secondary peaks or shoulders 5 to 8 h after oral ingestion (Julien-Larose et al., 1983; Grahnén et al., 1992), a phenomenon usually interpreted as indicative of enterohepatic cycling. Second, human feces contained the 10-N-glucuronide after oral administration of olanzapine (Kassahun et al., 1997) and a glucuronide of unidentified structure after ingestion of clozapine (Dain et al., 1997). These two tricyclic drugs bear a structural similarity to ketotifen, and in all three drugs the molecular masses of their glucuronides approach 500 Da, a value regarded as the threshold for efficient biliary excretion of amphiphilic substances into human bile (Klaassen and Watkins, 1984). Third, after i.v. administration of the quaternary amitriptyline N-glucuronide, deconjugation could be demonstrated in vivo (Breyer-Pfaff et al., 1990). It can, however, not be decided whether this took place in the intestine.

The N-glucuronidation kinetics of (S)-ketotifen in human liver microsomes were clearly biphasic, suggesting the participation of at least two enzymes with distinctly different affinities (Table 3, Fig. 6). In the presence of Triton X-100, the apparent K_M value of the high-affinity component was very similar to that found with amitriptyline as the substrate (Breyer-Pfaff et al., 1997), whereas that of the low-affinity component was lower. Two-enzyme kinetics could also be applied to an experiment without Triton addition. Under these conditions,(R)-ketotifen was N-glucuronidated biphasically, although the two components differed less inK_M (Table 4) and thus could not be as well evaluated as with (S)-ketotifen. Thus, it has to be concluded that as in the conjugation of amitriptyline, another member of the UGT enzyme family besides UGT1A4 is catalyzing the reaction. On Triton addition, (R)-ketotifen conjugation became monophasic in two of three microsomal preparations. A qualitatively similar behavior was described by Le Bigot et al. (1983), who studied the conjugation of racemic ketotifen and apparently measured the sum of all four N-glucuronides. In the absence of Triton, they found K_M values of 12.5 and 100 μM, the first one clearly exceedingK_M1 of 2.4 and 1 μM measured with(R)- and (S)-ketotifen, respectively, under the same conditions. Moreover, the V_max values of the previous authors were 2- to 4-fold lower. When Triton was present at 0.06% (3-fold the concentration used now), they observed monophasic glucuronidation with K_M 42 μM, a value that does not agree with any of the constants measured now. The detection of biphasic kinetics required the use of lowest drug concentrations in the low or submicromolar range in the present investigation and in a previous one (Breyer-Pfaff et al., 1997). The minimal concentration of racemic ketotifen used by Le Bigot et al. (1983) was 2 μM and those of tricyclic psychoactive drugs were 10 to 20 μM (Green et al., 1995, 1998; Green and Tephly, 1996). Still, the lowest ketotifen concentration of 0.5 μM was about 1000-fold higher than the maximal plasma level expected after a 1-mg dose (Grahnén et al., 1992). Therefore, the two enzymes catalyzingN-glucuronidation should contribute to total reaction rate according to their intrinsic clearances. If those measured with detergent in vitro correctly reflect the ratios in vivo (see below), the high-affinity enzyme is responsible for about 90% of(S)-ketotifen N-glucuronidation in liver, whereas no valid estimate can be made with regard to (R)-ketotifen. Some of the present experiments have been published in abstract form (Mey and Breyer-Pfaff, 1999), but the V_maxvalues reported are erroneous.

There were conspicuous differences between the contributions of individual ketotifen N-glucuronides to totalN-glucuronide production in vitro and to total excretion in human urine (Table 5). Intrinsic clearances derived from the ratios of apparent V_max andK_M values (Tables 3 and 4) should reflect the efficiencies of individual metabolic pathways. The majority of in vitro data was obtained in the presence of detergent because the latency of UGT enzymes that is overcome by the detergent is assumed not to prevail in vivo (Bossuyt and Blanckaert, 1995). On this basis,(R)-ketotifen glucuronides would be expected to account for about 40% of N-glucuronides formed from racemic ketotifen in human liver, whereas its contribution to N-glucuronides in urine was only 16%. In addition, the(S)-Gluc_B/(S)-Gluc_Aratio was 1.5 to 2 at low substrate concentrations in vitro, but on average 11 in urine. Several reasons are conceivable:N-Glucuronidation is possibly not confined to liver, but may be carried out in other organs with different stereoselectivity. For anO-glucuronidation this has been demonstrated, inasmuch as (+)-E-10-hydroxynortriptyline was conjugated in human liver microsomes and kidney homogenate, whereas the (−)-enantiomer was a UGT substrate in intestinal homogenate (Dahl-Puustinen et al., 1989). Alternatively, renal tubular secretion can be stereoselective, in this case with preferential transport of(S)-Gluc_B, whereas theN-glucuronides with lower renal clearances are more likely to undergo rehydrolysis. Tubular secretion of xenobiotic glucuronides has been claimed repeatedly (Møller and Sheikh, 1982), but firm experimental evidence from animal experiments is available only for acyl glucuronides (Meffin et al., 1983) and 1-naphthol glucuronide (Redegeld et al., 1988). Interestingly, morphine 6-glucuronide seemed to be reabsorbed in the rat kidney (van Crugten et al., 1991). If such a process occurs with ketotifen N-glucuronides, another possibility for stereoselectivity would be given. TheN-glucuronides may undergo enzymatic rehydrolysis either systemically or after biliary excretion (Sperker et al., 1997) to varying extents, as indicated by the experiments with rat and E. coli β-glucuronidase (Fig. 4).(S)-Gluc_B proved relatively resistant to the action of the bacterial enzyme, which should predominate in intestinal contents, but according to present knowledge reabsorption of unhydrolyzed glucuronide is rather improbable. Finally, competing routes of metabolism can affect the (R)/(S) ratio of N-glucuronides. A major reaction, reduction of the carbonyl group to the secondary alcohol, could be shown to be catalyzed by aldo-keto reductases in human liver cytosol, and these prefer(S)- over (R)-ketotifen as substrate (Breyer-Pfaff and Nill, 1999). Other reactions, namelyN-demethylation and N-oxidation, are of minor quantitative importance (Le Bigot et al., 1987) and their stereoselectivity is not known. The possibility of racemization in vivo has been checked by administering pure (R)-ketotifen to two volunteers. The low percentage of (S)-ketotifenN-glucuronides recovered from their urine would argue against a significant contribution of racemization to the observed discrepancy of (R)/(S) ratios of theN-glucuronides in vitro and in vivo.

In conclusion, conjugation of racemic ketotifen or one of its enantiomers at the tertiary amino group either chemically or enzymatically produced two quaternary ammonium glucuronides from each enantiomer. These are conformers differing in piperidylidene ring folding. The four isomers showed differential sensitivity toward enzymatic hydrolysis. The kinetics of their production in human liver microsomes were biphasic in the absence of a detergent and with(S)-ketotifen also in its presence, indicating the involvement of at least two UGT isozymes. Of an oral ketotifen dose, a mean of 17% was detected as N-glucuronides in urine with preferential excretion of one of the (S)-ketotifen glucuronides. Discrepancies between isomer ratios in microsomal incubates and in human urine can be due to extrahepaticN-glucuronidation, to differential rehydrolysis, and/or to selective transport of individual glucuronides.