Direct Characterization of Drug Glucuronide Isomers in Human Urine

Marianne Hädener , Wolfgang Weinmann , Dave R van Staveren , Stefan König .... John C. Lindon , Jeremy K. Nicholson , Jeremy R. Everett. 1999,1-88 ...
0 downloads 0 Views 343KB Size
Download PDF
Anal. Chem. 1996, 68, 2832-2837

Direct Characterization of Drug Glucuronide Isomers in Human Urine by HPLC-NMR Spectroscopy: Application to the Positional Isomers of 6,11-Dihydro-11-oxodibenz[b,e]oxepin-2-acetic Acid Glucuronide Eva M. Lenz,† David Greatbanks,‡ Ian D. Wilson,‡ Manfred Spraul,§ Martin Hofmann,§ Jeff Troke,⊥ John C. Lindon,† and Jeremy K. Nicholson*,†

Department of Chemistry, Birkbeck College, University of London, Gordon House, 29 Gordon Square, London WC1H 0PP, U.K., Department of Safety of Medicines, Zeneca Pharmaceuticals, Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, U.K., Bruker Analytische Messtechnik GmbH, Silberstreifen, D-76287 Rheinstetten, Germany, and Hoechst-Roussel, Walton Manor, Walton, Milton Keynes, Buckinghamshire MK7 7AJ, U.K.

In this work, 400 and 600 MHz 1H HPLC-NMR spectroscopic methods were developed and applied to separate and identify the positional glucuronide isomers and anomers of the model nonsteroidal antiinflammatory drug, 6,11-dihydro-11-oxodibenz[b,e]oxepin-2-acetic acid, in whole human urine. The HPLC methods utilized either an isocratic system, comprising 30% acetonitrile in water at pH 2.5, or a gradient elution system increasing from 30% to 60% acetonitrile, in order to achieve improved separation of the 2-, 3-, and 4-O-acylglucuronide isomers from the faster eluting endogenous urinary metabolites. Directly coupled stop-flow 1H HPLC-NMR spectroscopic measurements were made at the retention times indicated by the UV-monitored chromatographic peaks. The glucuronide isomers were identified from the 1H NMR spectra on the basis of their chemical shifts and spinspin coupling patterns. The elution order was 4-O-acyl-, 3-O-acyl-, and finally 2-O-acylglucuronide, with tR values of 10.04, 11.68, and 12.64 min, respectively. Although the r- and β-anomers of each of the positional isomers could not be separated in these solvent systems, they could be identified in the individual 1H NMR spectra. This work shows for the first time that directly coupled HPLCNMR spectroscopy can be used directly to isolate and characterize acyl-migrated isomers of drug glucuronides in whole urine. This approach will be of value in the study of glucuronide acyl migration reactions of nonsteroidal antiinflammatory drugs and other xenobiotic ester glucuronides in whole biofluids. β-Glucuronidation is a major in vivo metabolic pathway for many drugs and xenobiotics that contain hydroxyl and carboxylate groups. In the latter case, the resulting β-1-O-acylglucuronides can be highly susceptible to in vivo and in vitro acyl migration reactions at both physiological pH and under mild alkaline

Scheme 1

conditions, yielding a series of positional isomers and anomers. Such acyl migration reactions have been shown to occur with the glucuronide of 6,11-dihydro-11-oxodibenz[b,e]oxepin-2-acetic acid (I, Scheme 1). The acyl migration of the β-1-O-acylglucuronide (II) has been previously investigated in buffer solution by means of reversed-phase HPLC.1-3 It was found that the resulting acyl migration products, 2-, 3-, and 4-O-acylglucuronides, were unaf-

†

University of London. Zeneca Pharmaceuticals. § Bruker Analytische Messtechnik. ⊥ Hoechst-Roussel. ‡

2832 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

(1) Sidelmann, U. G.; Lenz, E. M.; Sanderson, P. N.; Spraul, M.; Hofmann, M.; Lindon, J. C.; Wilson, I. D.; Troke, J.; Nicholson, J. K. Anal. Chem. 1996, 68, 106-110. S0003-2700(95)01228-5 CCC: $12.00

© 1996 American Chemical Society

fected by incubation with the specific enzyme, β-glucuronidase,2,3 as has been reported for other transacylated glucuronide isomers.4-6,8 It has also been shown that the rate of acyl migration of drug glucuronides is highly dependent on the pH of the matrix2,3,5 and that acidification of urine containing the glucuronide isomers halts the acyl migration reactions.2,3 Under alkaline pH conditions, acyl migration and hydrolysis are competing reactions, hydrolysis predominating at pH > 10.2,3 In addition to the pH dependence of the acyl migration rates, the structural properties of the drug itself exert a strong influence on the observed acyl migration rates in vitro and in vivo, as has been shown for the glucuronides of several proprietary drugs.8 The relative contributions of the underlying physicochemical, steric, and electronic properties that control acyl migration rates under physiological conditions are, however, poorly understood. The products of intramolecular rearrangement reactions of ester glucuronides have been observed for a wide range of compounds, including clofibric acid,9 bilirubin,7 probenecid,10 and substituted benzoic acids.11-13 The relative propensity of drug glucuronide conjugates to undergo acyl migration may be associated with allergic responses in humans, since the rearranged glucuronide isomers can react with serum and cellular nucleophiles (e.g., lysyl amine groups in albumin).8 There is, therefore, a need to develop novel analytical approaches for the investigation of acyl migration reactions of drug glucuronides so that structure-reactivity relationships can be constructed with a long-term aim of effecting drug design to minimize the toxicological risks associated with reactive glucuronide formation. The recent development of high-frequency directly coupled HPLC-NMR spectroscopy has led to a number of applications in the analysis of drugs and drug metabolites in biological fluids.14-16 We have also investigated the utility of HPLC-NMR methods for the separation and structural characterization of isomeric ester glucuronides of model compounds present in equilibrium mixtures formed in acyl migration reactions from the β-1-O-acylglucuronide.1,11-13 We have recently shown that 750 (2) Paul, H.; Illing, A.; Wilson, I. D. Biochem. Pharmacol. 1981, 30, 33813384. (3) Wilson, I. D.; Bhatti, A.; Illing, H. P. A.; Bryce, T. A.; Chamberlain, J. In Drug Metabolite Isolation and Determination; Reid, E., Leppard, J. P., Eds.; Plenum Press: New York, 1983; pp 181-187. (4) Marsh, C. A. In Glucuronic Acid Free and Combined; Dutton, G. J., Ed.; Academic Press: New York, 1966; pp 185-201. (5) King, A. R.; Dickinson, R. G. Biochem. Pharmacol. 1991, 42, 2289-2299. (6) Caldwell, A.; Hutt, A. J.; Marsh, M. V.; Sinclair, K. In Drug Metabolite Isolation and Determination; Reid, E., Leppard, J. P., Eds.; Plenum Press: New York, 1983; p 161-179. (7) Compernolle, F.; Van Hees, G. P.; Blanckaert, N.; Heirwegh, K. P. M. Biochem. J. 1978, 171, 185-201. (8) Spahn-Langguth, H.; Benet, L. Z. Drug Metab. Rev. 1992, 24, 5-48. (9) Faed, E. M.; McQueen, E. G. Clin. Exp. Pharmacol. Physiol. 1978, 5, 195198. (10) Upton, G. A.; Buskin, J. N.; Williams, R. Z.; Holford, N. H. G.; Riegelman, S. J. Chromatogr. 1980, 145, 393-40. (11) Sidelmann, U. G.; Gavaghan, C.; Carless, H. A. J.; Farrant, R. D.; Lindon, J. C.; Hansen, S. H.; Wilson, I. D.; Nicholson, J. K. Anal. Chem., in press. (12) Sidelmann, U. G.; Nicholls, A. W.; Meadows, P. E.; Gilbert, J. W.; Lindon, J. C.; Wilson, I. D.; Nicholson, J. K. J. Chromatogr. 1996, 728, 377-385. (13) Sidelmann, U. G.; Gavaghan, C.; Carless, H. A. J.; Spraul, M.; Hofmann, M.; Lindon, J. C.; Wilson, I. D.; Nicholson, J. K. Anal. Chem. 1995, 67, 34013404. (14) Spraul, M.; Hofmann, M.; Wilson, I. D.; Lenz, E. M.; Nicholson, J. K.; Lindon, J. C. J. Pharm. Biomed. Anal. 1993, 11, 1009-1015. (15) Wilson, I. D.; Nicholson, J. K.; Spraul, M.; Hofmann, M.; Lindon, J. C. J. Chromatogr. 1993, 617, 324-328. (16) Spraul, M.; Hofmann, M.; Dvortsak, P.; Nicholson, J. K.; Wilson, I. D. Anal. Chem. 1993, 65, 327-330.

MHz HPLC-NMR spectroscopy can be used to separate and identify the individual glucuronide isomers (both positional and R- and β-anomers) of II in equilibrium mixtures in buffer solutions using a pH 7.4 phosphate buffer/acetonitrile eluent.1 However, real biological fluids, such as urine and plasma, present much greater analytical challenges than simple buffer solutions, as, in addition to the presence of a large number of matrix components, the various glucuronide positional isomer concentrations can be low. In the present study, we show the successful application of 400 and 600 MHz stop-flow HPLC-NMR to the detection and characterization of positional isomers and anomers of II present in whole human urine injected directly onto the HPLC column. EXPERIMENTAL SECTION Dosing and Urine Collection. An adult male volunteer (70 kg weight) received a single 250 mg oral dose of I. Urine samples were collected over 0-1.5, 1.5-3, and 3-5 h time periods and acidified to pH 3.5 immediately on collection. Samples were stored frozen at -20 °C until analysis. A portion of the 1.5-3 h urine sample was divided into three aliquots. The first was adjusted to pH 10 to hydrolyze the 1-O-acylglucuronide and liberate the free aglycon. This sample was then acidified for subsequent HPLCNMR analysis. The second aliquot was kept at pH 3.5 and was also subjected to HPLC-NMR to confirm the presence of the drug 1-O-acylglucuronide in the urine sample. The third sample was adjusted to pH 8 and left to equilibrate at room temperature overnight to promote the formation of 2-, 3-, and 4-O-acylglucuronides. The sample was then acidified to pH 2 before analysis by HPLC-NMR. HPLC Solvents. HPLC solvents were purchased from Riedel de Haen, Germany (acetonitrile, ACN), and Fluorochem, U.K. (D2O), and were both of analytical grade. 400 MHz HPLC-NMR Spectroscopy. Chromatography was initially carried out on a 25 cm × 4.6 mm i.d., column packed with 5 µm Spherisorb ODS-2 (Phase Separations Ltd., U.K.). The mobile phase consisted of 34% ACN/66% H2O (v/v) containing 1% concentrated formic acid to give a final pH of 2.5 (in order to suppress ionization of the analytes). For ease of solvent resonance suppression, the water was replaced by D2O (99.9%) for HPLCNMR. This is standard practice, given the relatively low cost of D2O. The chromatographic separation was optimized so that all endogenous urinary components eluted before the glucuronide isomers. For the elution of the β-1-O-acylglucuronide (II) and the aglycon (I), the mobile phase was maintained at 34% ACN. However, the ACN content was lowered to 30% to achieve better separation of the isomeric glucuronides during the HPLC-NMR analysis. The HPLC-NMR separation was performed isocratically at a flow rate of 1 mL/min with a Bruker LC22C pump using a 200 µL injection. The eluent from the column was monitored using a Bruker LC33 UV detector at 280 nm. 1H HPLC-NMR analyses were carried out on a Bruker AMX 400 spectrometer operating at a 400.13 MHz 1H resonance frequency in the “stop-flow” mode. HPLC-NMR experiments were performed on the basis of the chromatographic retention times of the analyte peaks. Double solvent suppression of the ACN and the residual water in the D2O mobile phase was achieved by applying the standard NOESYPRESAT pulse sequence (UXNMR, Bruker Spectrospin Ltd.).17 This is simply the first (17) Nicholson, J. K.; Foxall, P. J. D.; Spraul, M.; Farrant, R. D.; Lindon, J. C. Anal. Chem. 1995, 67, 793-811.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2833

increment of the NOESY sequence, with a delay between the first two 90° pulses of 3 µs and with irradiation of the water resonance for 1.4 s before the application of the first pulse and during the 80 ms mixing time. For the 400 MHz NMR spectra, the total number of scans was varied according to the concentration of the fractions (from 160 to 5000). Spectra were acquired over a 8000 Hz spectral width into 16K data points. The flow cell volume was 120 mL, resulting in a total transfer time from the UV detector of 32 s. The acquisition time was 1.02 s. A relaxation delay of 1.45 s was included. Spectra were zero-filled by a factor of 2 prior to Fourier transformation. 600 MHz HPLC-NMR Spectroscopy. The chromatography was carried out with a J’sphere ODS H80 (YMC, Japan) 4 mm column (25 cm × 4.6 mm i.d.). The separation was performed at a flow rate of 0.8 mL/min with a Shimadzu 10A pump and autosampler. A linear gradient was applied after 10 min of isocratic flow with 30% ACN/D2O (acidified with 0.05 M H3PO4), which increased the ACN concentration to 60% over 10 min. An aliquot of the equilibrated urine was freeze-dried and reconstituted in the initial mobile phase concentration at an 8-fold concentration (5 mL/600 mL of 30% ACN), of which a 50 mL aliquot was injected onto the column. Therefore, the concentration of the urine sample was doubled relative to the 400 MHz HPLC-NMR analysis. 600 MHz HPLC-NMR spectroscopy was carried out on a Bruker DRX600 spectrometer at 600.14 MHz under the same chromatographic conditions as those described above, using a flow cell of 4 mm i.d. (120 µL). Spectra were measured with 320 scans, with a spectral width of 12 000 Hz, and data were acquired into 32K data points. An acquisition time of 1.37 s and a relaxation delay of 1.83 s were used. No zero-filling was applied. RESULTS AND DISCUSSION For the chromatographic analysis of the principal urinary drug metabolite, the β-1-O-acylglucuronide (II), a mobile phase concentration of 34% ACN in D2O was used, and the HPLC-UV chromatogram is shown in Figure 1a. The 1H 400 MHz HPLCNMR spectrum obtained by stopping the flow at the retention time of the β-1-glucuronide (tR ) 5.6 min) is shown in Figure 2. This spectrum shows clearly the characteristic doublet at δ 5.5 due to the anomeric proton of the β-1-O-acylglucuronide conjugate and the other glucuronide ring proton signals in the 1H chemical shift range δ 3.4-4.0. The 1H NMR resonances from the aromatic protons and the CH2(a) group can also be assigned as shown in Figure 2. The UV-detected HPLC chromatogram of the urine after alkaline hydrolysis to liberate the aglycon I is shown in Figure 1b. This shows an increase in I and a decrease in II. The chemical shifts are summarized in Table 1. Following incubation of the urine sample as described in the Experimental Section, the 1-O-acylglucuronide underwent acyl migration reactions to produce the isomers as indicated in Scheme 1. Stop-flow 400 MHz 1H HPLC-NMR spectra of the glucuronide isomers were measured at points in the chromatogram indicated by their absorption peaks from the in-line UV detector. The UV chromatogram obtained for the separation of the glucuronide isomers, from 200 µL of the equilibrated and acidified urine, with an eluent of 30% ACN is shown in Figure 3. The use of 1H 400 MHz stop-flow HPLC-NMR spectroscopy enabled the direct identification of the 2-, 3-, and 4-positional glucuronide isomers. The R- and β-anomers of each of these isomers were poorly chromatographically resolved, but each anomer could still be readily identified by their characteristic 1H chemical shifts and 2834 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

a

b

Figure 1. (a) HPLC-UV chromatogram of human urine, showing presence of the glucuronide metabolite II, obtained using 34% ACN/ D2O as eluent. (b) HPLC-UV chromatogram of human urine after alkaline hydrolysis, showing an increased level of the aglycon I, obtained using the same eluent.

spin-spin coupling patterns. Attempts to improve chromatographic resolution by lowering the ACN concentration of the eluent further were unsuccessful (data not shown). In previous HPLC-NMR studies on fluorobenzoic acid glucuronide isomers and anomers, the chromatographic resolution of R- and β-anomers was less problematic.18 A number of minor HPLC solvent impurities (methanol, 2-propanol, propanol) were detected in our systems using 1H NMR spectroscopy, and these gave resonances in the same region as the glucuronide ring. These impurities appear to be common contaminants of HPLC-grade solvents, which do not normally interfere with conventional HPLC analyses because they have very weak UV chromophores. However, in HPLC-NMR studies, particularly those involving low concentrations of analytes, the minor solvent impurities can cause significant (18) Sidelmann, U. G.; Gavaghan, C.; Carless, H. A. J.; Spraul, M.; Hofmann, M.; Lindon, J. C.; Wilson, I. D.; Nicholson, J. K. Anal. Chem. 1995, 67, 44414445.

Figure 2. Partial 400 MHz 1H HPLC-NMR spectrum of the chromatographic peak corresponding to (II) acquired under stop-flow conditions after 160 transients and Lorentzian-Gaussian resolution enhancement. The singlet at δ 3.3 has been identified as methanol impurity in one batch of ACN. Table 1. 1H HPLC-NMR Chemical Shifts (δ) for the Glucuronide Ring Protons of the Positional Isomers of 6,11-Dihydro-11-oxodibenz[b,e]oxepin-2-acetic Acid Glucuronide (II) Obtained by Stop-Flow HPLC-NMR Analysisa tR β-1-O-acyl R-4-O-acyl β-4-O-acyl R-3-O-acyl β-3-O-acyl R-3-O-acyl β-3-O-acyl R-2-O-acyl β-2-O-acyl

5.6 10.04 10.04 11.68 11.68 17.1 17.1 12.64 12.64

NMR freq 400 400 400 400 400 600 600 400 400

1′

2′

3′

4′

5′

5.5 d 5.18 d 4.58 d 5.18 d 4.65 d 5.21 d 4.68 d 5.29 d 4.71 d

3.4-3.5c

3.4-3.5c

3.4-3.5c

3.53 d 3.43 d 3.63 t 3.35 t 3.65 dd 3.37 dd 4.63 dd b

3.84 t 3.65 t 5.10 t 4.93 t 5.13 t 4.95 t 3.85c 3.85c

4.84 t 4.83 t b 3.67c 3.69 t 3.68 t 3.59 t 3.58 t

3.95 d 4.35 d 4.04 d 4.29 d 3.98 d 4.30 d 3.97 d 4.25 d 3.97 d

a

tR refers to the retention time (min) in the HPLC separation. NMR freq is the NMR observation frequency (MHz). d, doublet; dd, doublet of doublets; t, pseudotriplet. Typical axial-axial and axial-equatorial coupling constants were 9-10 Hz and 3-4 Hz, respectively. b 1H NMR signal not detected due to overlap or poor signal-noise ratio. c Peak splittings incompletely resolved.

interferences if the chemical shifts are close to these of the compounds of interest. The assignments of the NMR spectra from the various chromatographic peaks to the 2-O-acyl, 3-O-acyl, and 4-O-acyl isomers was based on the known chemical shifts of the protons in both R- and β-glucuronic acid, as determined by spin decoupling experiments (results not shown), and known effects induced by acylation on 1H chemical shifts. The first glucuronide peak eluted after 10.04 min and was due to the 4-O-acylglucuronide isomers (VII, VIII) as a mixture of Rand β-anomers, as shown by the 1H NMR spectrum (Figure 4). At δ 4.83, two overlapped “pseudotriplets” (each an uncompletely resolved doublet of doublets) were observed, corresponding to

Figure 3. Reversed-phase HPLC-UV chromatogram of urine following incubation at pH 8 with 30% ACN and showing the presence of the 2-, 3-, and 4-O-acyl positional isomers. The acyl migration products eluted with retention times of 12.64, 11.68, and 10.04 min, respectively.

the H4′ proton on the glucuronide ring. The change from the corresponding proton chemical shift in the 1-O-acylglucuronide to high frequency by approximately 1-1.15 ppm appears to be characteristic for secondary alcohols and sugars following acylation,19 with additional deshielding due to the acidic medium, resulting in protonation of the glucuronide carboxylate group. The occurrence of two pseudotriplets with slightly different chemical shifts stems from the mutarotation of the deacylated OH group at the C1′ position, resulting in nearly equimolar proportions of the R- and β-anomers of the 4-O-acylglucuronide. A further consequence of the deacylation of the C1′ hydroxyl group is the low-frequency shift of the resonance of the H1′ proton of the (19) Casey, A. F. PMR Spectroscopy in Medicinal and Biological Chemistry; Academic Press: London, 1971; p 30.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2835

Figure 4. Partial 400 MHz 1H NMR spectrum of the peak eluting at 10.04 min, representing the mixture of R- and β-anomers of the 4-Oacylglucuronide (VII). The spectrum was aquired after 968 transients, and Lorentzian-Gaussian resolution enhancement was used.

β-anomer from δ 5.5 to δ 4.58. The H1′ proton resonance of the R-anomer appears at δ 5.18. At δ 4.04 and δ 4.35, two doublets are observed from the H5′ signal of the R- and β-anomers. However, the relative intensities of these doublets are distorted due to partial saturation of the signal at δ 4.35 because of the water-suppression irradiation applied at about δ 4.5. The H3′ signals appear as two sets of pseudotriplets at δ 3.84 and δ 3.65 for the R- and β-forms respectively. Two overlapped singlets from the alkyl side chain proton of the aglycon moiety (labeled a) are visible at δ 3.74, again from the R- and β-glucuronide anomers. The remaining signals at δ 3.53 and δ 3.43, a doublet of doublets and a pseudotriplet, respectively, derive from the H2′ proton of the glucuronide. The different coupling patterns are the result of the differences in the orientation of the 1-OH group of the Rand β-anomers; i.e. the H2′ couples to an axial H3′ and an axial H1′ when the glucuronide is in the β-form, giving rise to a pseudotriplet. However, with the R-glucuronide, the H2′ couples to an axial H3′ proton and an equatorial H1′ proton, resulting in the characteristic doublet of doublets pattern. The two chromatographic peaks could not be resolved further by lowering the ACN content of the mobile phase (data not shown), indicating that the mutarotation rate of the 4-O-acylglucuronide is rapid relative to the time involved in the chromatographic separation but still slow on the NMR time scale. The peak eluting at 11.68 min corresponds to the 3-Oacylglucuronides (V and VI). This chromatographic peak has been examined using both 400 and 600 MHz HPLC-NMR 2836 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

Figure 5. Partial 600 MHz 1H NMR spectrum of the mixture of Rand β-anomers of the 3-O-acylglucuronide (V), eluting at 17.1 min after lyophilization and 2-fold concentration. The spectrum was acquired after 320 transients, and no FID weighting function was used.

spectroscopy. In the latter case, the altered chromatographic conditions as detailed in the Experimental Section resulted in a longer retention time of 17.1 min. Again, mutarotation is observed due to lack of acylation at the C1′ position of the glucuronide. The proportions of the anomers in the 400 MHz NMR study were nonequal, however, with a considerable excess of the β-anomer. In a separate HPLC-NMR study at 600 MHz, this chromatographic peak showed the presence of both the R- and β-anomers of the 3-O-acyl positional isomer in approximately equimolar concentrations (Figure 5). The doublet resonances from H5′ were deshielded from their positions in the 1-O-acylglucuronide and were of almost equal signal intensity, the doublet closer to the water resonance (from the R-anomer at δ 4.30) being slightly suppressed because of radio frequency irradiation at the frequency of the solvent resonance. The signal from the β-anomer appears at δ 3.97. The doublet resonances from the H1′ protons appear, as expected, at characteristic chemical shifts of δ 5.21 and δ 4.68 for the R- and β-anomers, respectively. The signals arising from the H2′ protons can be fully accounted for with the doublet of doublets pattern due to H2′ in the R-anomer appearing at δ 3.65 and the more shielded “triplet” (visible as a doublet of doublets) due to H2′ in the β-anomer at δ 3.37. Signals for the H4′ protons appear overlapped at δ 3.68, and the triplet resonances from the H3′ protons appear well separated, as expected from reference shifts of glucuronic acid itself at δ 5.13 and δ 4.95. The methylene side chain signal at δ 3.81 shows the characteristic doubling due to the presence of both anomeric forms.

In the 400 MHz 1H HPLC-NMR spectra of the peak eluting at 12.64 min, the R- and β-anomers of the 2-O-acyl isomers (III and IV) can be seen. The limited chromatographic resolution resulted in incomplete separation of the 2-O-acyl and 3-O-acyl isomers, and evidence for the latter can be seen in the spectra of the fraction eluting at 12.64 min. This, however, did not preclude NMR signal assignment. The doublet at δ 5.29 is assigned to the H1′ proton of the R-anomer of the 2-O-acyl isomer, and the corresponding signal due to the β-anomer is also a doublet at δ 4.71. The H2′ proton of the R-anomer of the 2-O-acyl isomer shows the characteristic doublet of doublets pattern due to coupling to the axial H3′ and the equatorial H1′ protons but is overlapped with the doublet derived from the H1′ proton of the 3-O-acyl β-anomer present in that chromatographic fraction. The corresponding signal for the β-anomer of the 2-O-acyl isomer is obscured by the solvent signal at δ 4.5. In this case, there is a slight excess of the R-anomer. This can be shown by the anomeric doublets and the signal of the alkyl side chain singlet at δ 3.79. The doublets for the H5′ proton are visible at δ 4.25 and δ 3.97 for the R- and β-anomers respectively. The residual H3′ proton triplets resonate at δ 3.85, and the H4′ protons at δ 3.58. There were no 1-O-acylglucuronide NMR signals visible in this last chromatographic fraction, even though its retention time coincides with the 2-O-acylglucuronide peak. Although the ability to separate the 2-, 3-, and 4-O-acyl transacylation products from urine is clearly demonstrated, there are limitations in separating the R- and β-anomers of this model drug glucuronide, since mutarotation occurs rapidly over a wide pH range,5 and hence, this happens to a certain extent during the HPLC-NMR analysis itself. The 400 MHz HPLC-NMR system also appeared to be adequate for most of the glucuronide characterization in this study, although 600 MHz HPLC-NMR was used in this case to provide better sensitivity and dispersion for one of the isomeric glucuronides. This study shows the first example of the separation of isomeric drug glucuronides from a complex biological matrix (urine) by directly coupled HPLC-NMR spectroscopy. The method was successfully applied to the identification of the acyl migration products of the glucuronides of 6,11-dihydro-11-oxodibenz[b,e]oxepin-2-acetic acid (I) without the need for enzymic

hydrolysis. The glucuronide separations were carried out with virtually no sample preparation, although the equilibration of glucuronide isomers was stimulated by raising the pH of the original urine sample with reacidification prior to HPLC injection. Nevertheless, the 1H HPLC-NMR spectra could be confidently attributed to each of the individual transacylation products. Only one chromatographic peak (that of the 3-O-acylglucuronide isomer) required reanalysis at higher NMR observation frequency in order to fully characterize both anomeric forms. It is noted that the chromatographic elution order of the acylmigrated isomers determined here was the same (4-O-acyl-, 3-Oacyl-, and 2-O-acylglucuronides) as that used previously for the acyl migration products in acetonitrile/D2O/0.2 M KH2PO4 (21: 69:10) at pH 7.4.1 This indicates that there is a standard order of elution for drug glucuronides, irrespective of R group over a wide pH range. Most importantly, this work shows the practical utility of HPLC-NMR spectroscopy in the characterization of drug glucuronide isomers in whole biofluids. Clearly, in future studies, the HPLC conditions will require individual optimization for the separation of the glucuronide isomers of each new drug to be investigated. However, it is likely that this will always be achievable, at least for the positional isomers, with simple water/ acetonitrile mixtures, as shown in this study and in previous studies.12,18 The HPLC-NMR approach is now suitably developed for application to a wider range of investigations into drug glucuronide acyl migrations in biofluids and the study of structurereactivity relationships for these metabolites, which will be of benefit to the understanding of some types of adverse drug reactions in humans. ACKNOWLEDGMENT We thank the EPSRC for supporting this project and Zeneca Pharmaceuticals and the Wellcome Research Laboratories for a CASE studentship. Received for review December 19, 1995. Accepted May 17, 1996.X AC951228L X

Abstract published in Advance ACS Abstracts, July 1, 1996.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2837