Introduction

Top Abstract Introduction Results Discussion References

Rhinoviruses are the primary causative agent implicated in human upper respiratory tract infections and are responsible for a large percentage of common cold incidents each year (Couch, 1990; Rueckert 1990; Gwaltney and Rueckert, 1997). The production of mature virion particles requires a functional rhinovirus 3C protease, which is responsible for post-translational cleavage of viral polyprotein precursor into active enzymatic proteins (Krausslich and Wimmer, 1988; Rueckert, 1990). The importance of this enzyme in viral replication has led to its identification as a novel therapeutic target (e.g., Kaldor et al., 1995; Webber et al., 1996; Dragovich et al., 1998a,b; Kong et al., 1998; Wang et al., 1998).

AG7088 [trans-(4S,2'R,5'S,3S)-4-{2'-4-(4-fluorobenzyl)-6'-methyl-5'-[(5"-methylisoxazole-3"-carbonylamino]-4-oxoheptanoylamino}-5-(2-oxopyrrolidin-3--yl)pent-2-enoic acid ethyl ester; Fig. 1] is a potent and irreversible peptidomimetic inhibitor of rhinovirus 3C protease (K_obs/I = 1.5 × 10⁶ M¹ s¹) that was discovered using rational protein structure-based drug design technology (Dragovich et al., 1998a,b, 1999a,b). In H1-HeLa and MRC-5 cell protection assays, AG7088 was shown to possess a broad spectrum of antiviral activity, with ED₉₀ values ranging from 0.018 to 0.261 µM against 48 rhinovirus serotypes (Patick et al., 1999). Despite this demonstration of broad in vitro activity, the development of an appropriate in vivo efficacy model presents significant challenges. Among all animals evaluated, the chimpanzee is the only species that can harbor the rhinovirus; however, it does not exhibit any of the symptoms associated with human rhinoviral infection (Dick, 1968). In the absence of an animal efficacy model for rhinovirus infection, selection of appropriate animal models for safety evaluation partly relies upon exposure to drug and drug-derived metabolites most similar to that present in humans. Previous in vitro metabolic studies have shown that AG7088 undergoes extensive metabolism in hepatic microsomes from various species (Harr et al., 1999). The current study describes and compares the potential sites of metabolism of AG7088 in various animal species using liver microsomes isolated from the mouse, rat, rabbit, dog, monkey, and human, with the aim of using the results in the selection of the appropriate animal models for safety evaluation. The microsomal metabolite identification was performed using complementary hyphenated techniques, LC-MS/MS³ and LC-NMR.

View larger version (13K): [in this window] [in a new window]   Fig. 1.   Chemical structure of AG7088, which comprises four main building blocks denoted P1/P1', P2, P3, and P4.

Experimental Procedures

Materials. AG7088 and AG7185 (M4) were synthesized at Agouron Pharmaceuticals, Inc. (San Diego, CA). NADPH, monobasic potassium phosphate, and dibasic potassium phosphate were purchased from Sigma Chemical Co. (St. Louis, MO). Dimethyl sulfoxide and trifluoroacetic acid (spectrophotometric grade) were purchased from Aldrich Chemical (Milwaukee, WI). Male mouse and rat liver tissues were purchased from Harlan Bioproducts for Science, Inc. (Indianapolis, IN). Male rabbit, dog, and monkey liver tissues were obtained from HRP-Covance (Denver, PA). Human liver tissues were purchased from the International Institute for Advancement of Medicine (Exton, PA). Additional dog liver microsomes (male beagle) were obtained from In Vitro Technologies (Baltimore, MD). Acetonitrile was obtained from Fisher Scientific (Pittsburgh, PA; HPLC grade) or Aldrich [Riedel-deHaen; NMR Chromasolv (LC-NMR grade)]. Deuterium oxide (99.9% D) was purchased from Cambridge Isotopes.

Preparation of AG7088 Stock solutions for Microsomal Incubations. A 50 mM solution of AG7088 in dimethyl sulfoxide was initially prepared and subsequently diluted with acetonitrile to achieve a 1 mM solution.

AG7088 Incubations in Various Animal and Human Liver Microsomes. Liver microsomes (1 mg of protein/ml) were incubated with AG7088 (25 µM) in 100 mM potassium phosphate, pH 7.4, at 37°C in a shaking water bath for 5 min before initiating the reaction with NADPH (2 mM). The final volume of each incubation was 0.5 ml. After the addition of NADPH, a 200-µl aliquot was removed and placed into a test tube containing 2 ml of acetonitrile to serve as a time 0 sample. The remainder of the incubate proceeded for 30 min, after which the reaction was terminated by the addition of acetonitrile (2 ml). All samples were vortexed for 2 min on an SP Multi-tube Vortexer (Baxter, McGaw Park, IL) and then centrifuged at 2500g for 20 min. The organic layer was transferred to another set of test tubes, and the solvent was evaporated under a gentle stream of nitrogen using a Dri-Block sample concentrator (Techne, Princeton, NJ) at 50°C. The sample residues were stored at 20°C until analyzed by LC-MS/MS.

LC-MS/MS. AG7088 metabolites produced by small-scale incubates were analyzed by LC-MS/MS. Interpretation of the product ion mass spectra (MS/MS) of each metabolite was conducted in reference to the parent drug. The LC-MS/MS system included a Waters 2690 Separation Module (Waters, Milford, MA) and a Quattro II triple stage quadrupole mass spectrometer controlled by Masslynx 3.0 software (Micromass, Beverly, MA). The chromatography was performed on a reversed phase column (Waters Symmetry C₁₈, 2.1 × 150 mm, 5 µm) using a gradient elution method at a flow rate of 0.2 ml/min. The mobile phase contained 0.01% TFA in water (A) and 0.01% TFA in acetonitrile (B). Gradient elution was programmed linearly from 20 to 80% B over 20 min. The mass spectrometer was operated under the following conditions: electrospray in positive ion mode using a Crossflow counter electrode, capillary voltage = 3.2 kV, cone voltage = 30 V, source temperature = 140°C, collision energy = 22 eV, and collision gas cell pressure = 1.4 × 10³ mBar. Data were collected by MS1 under full scan mode (m/z 150-800 at a scan speed of 1.30 s), and daughter ion scan mode to obtain structural information.

LC-NMR. AG7088 and metabolites M1, M2, M3, and M4 were analyzed by LC-NMR on a Bruker Avance 500-MHz spectrometer (Bruker Instruments, Fremont, CA). The system was configured with a 4-mm ¹H/¹³C inverse-geometry LC (LC SEI) probe equipped with x,y,z-gradients, an HP1100 analytical HPLC with binary pump and variable wavelength UV detector, and a Bruker 12-loop peak sampling unit (BPSU-12). The LC-NMR software interface was HyStar NT Version 1.2 (Bruker Instruments). Chromatography was performed on a reversed phase column (Capcell C₁₈, 4.6 × 250 mm, 5 µm) using a gradient elution method at a flow rate of 1 ml/min. Mobile phase A contained 0.1% TFA in deuterium oxide (99.9% D), while mobile phase B contained 0.1% TFA in acetonitrile. The solvent gradient was increased linearly from 20% B to 50% B over 30 min. The UV absorbance was monitored at 220 nm. LC-NMR was performed in stop-flow mode. AG7185 (M4) eluted at 41% mobile phase B with a retention time (t_R) of 21.6 min; M3 at 45% B with t_R = 25.1 min, M1/M2 at 47% B with t_R = 27.3 min, and AG7088 at 50% B with t_R = 30.4 min. Solvent suppression was performed using either double presaturation of the residual solvent resonances, or WET (Ogg et al., 1994; Smallcombe et al., 1995) solvent suppression with ¹³C decoupling. Spectra were acquired using 64K data points and a spectral width of 10,000 Hz. Typically, data were acquired overnight (7K-20K transients). In addition, reference LC-NMR spectra were acquired on synthetic standards of AG7185 and AG7088 (50-µg quantities with retention times of 21.5 and 30.0 min, respectively), and one-dimensional spectra were acquired with 1K transients. To obtain ¹H resonance assignments for AG7088 under the LC solvent conditions, magic angle gradient double quantum-filtered correlation spectroscopy (van Zijl et al., 1995) was performed on AG7088 synthetic standard (200 µg of AG7088, t_R =30.3 min). The two-dimensional data were collected in phase-sensitive mode using TPPI, and gradient pulses were employed for coherence transfer selection. For each of the 256 t₁ increments, 704 transients were acquired using 2K data points over a spectral width of 7000 Hz, with the carrier centered on the acetonitrile resonance. All spectral data were referenced to acetonitrile at 2.0 ppm.

	Results

Top Abstract Introduction Results Discussion References

HPLC Profiles of AG7088 Metabolites. The HPLC chromatographic profiles of AG7088 and metabolites produced by hepatic microsomes from the mouse, rat, rabbit, dog, monkey, and human are illustrated in Fig. 2, a to f, respectively. The metabolites are designated M1 to M6 on the basis of increasing polarity or decreasing retention time on a reversed phase HPLC column. M4 (AG7185) was the predominant metabolite in all six species studied, especially in rodents and rabbits. Metabolites M1 to M3 were more prevalent in human and dog liver microsomes, whereas M5 and M6 were more abundant in monkey liver microsomes. The abundance of each metabolite was graded from 1+ to 5+ (Table 1) based on its peak height relative to AG7088 (Fig. 2, a-f). Results for M1 and M2 were pooled (vide infra).

View larger version (20K): [in this window] [in a new window]   Fig. 2.   HPLC chromatograms (analog UV signal at 220 nm) of AG7088 metabolites formed by liver microsomes in various species [mouse (a), rat (b), rabbit (c), dog (d), monkey (e), and human (f)].

Structural Elucidation by LC-MS/MS and LC-NMR.

Parent drug AG7088 (Fig. 1) comprises four main building blocks: P1, a Glu-derived -lactam containing an ethyl propenoate Michael-acceptor moiety (P1'); P2, 4-fluoro-Phe; P3, Val, which is joined to P2 via a ketomethylene isostere; and P4, a methyl isoxazole (Dragovich et al., 1998a,b, 1999a,b). Identification of the microsomal metabolites of AG7088 was accomplished using LC-MS/MS and LC-NMR and was facilitated by comparative analysis with the parent compound. The product ion mass spectrum of AG7088 (MH⁺ = 599; Fig. 3a) provided structural information mainly from the fragmentation at the P1 to P2 linkage (m/z 210, 227, 345, and 373). In addition, fragmentation at the P3 to P4 linkage produced an ion at m/z 455. In the LC-NMR spectrum of AG7088, key functional groups were clearly represented. These included the fluorophenyl ( 7.15, dd, 2H;  6.97, dd, 2H) and methyl isoxazole ( 6.45, s, 1H;  2.44 s, 3H) rings, the trans-,-unsaturated ethyl ester ( 6.53, dd, 1H;  5.23, d, 1H;  4.14, q;  1.25, t, 3H), the valine methyl and C protons ( 0.94, d, 3H;  0.82, d, 3H;  4.55, d, 1H), and the P1 lactam protons adjacent to the nitrogen ( 3.25, 3.18, m), as well as several aliphatic resonances associated with P2 ( 2.613.10).

View larger version (25K): [in this window] [in a new window]   Fig. 3.   Product ion mass spectra of AG7088 (a) and its metabolites [M1/M2 (b); M3 (c); M4 (d); M5 (e); and M6 (f)], and assignment of fragment ions.

Metabolites M1 and M2. The MS1 full scans of metabolites M1 and M2 showed that both peaks contained the ion at m/z 615 as the most abundant ion, consistent with monohydroxy metabolites of AG7088. Their MS/MS spectra were almost identical, in which the protonated molecule and most of the fragment ions containing the P1/P1' moiety readily lost a water molecule (e.g., m/z 208, 225, 453, and 597; Fig. 3b), presumably to form thermodynamically more stable or fully conjugated products. In contrast, the fragment ions that did not contain the P1/P1' moiety remained unchanged (e.g., m/z 345 and 373). This mass spectral information localized the hydroxy groups of M1 and M2 to the P1/P1' moiety. However, it was not possible to definitively assign the structures from these data alone, and M1 and M2 were further characterized by LC-NMR.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   LC-NMR spectra of AG7088 (top; R = H) and metabolites M1/M2 (bottom; R = OH), expanded about the region from 4.45 to 7.40 ppm.

Metabolite M3. The MS1 full scan spectrum of M3 also showed the ion at m/z 615 as the most abundant ion, consistent with a monohydroxy metabolite of AG7088. In contrast to M1 and M2, M3 retained all fragment ions that included the P1/P1' moiety (e.g., m/z 210, 227, and 455), but shifted +16 Da on ions that contained the P4 moiety (e.g., m/z 143, 361, and 389). These data (Fig. 3c) localized the M3 hydroxy group to the methylisoxazole ring. LC-NMR definitively established the site of oxidation as the P4 methyl group. The AG7088 methyl singlet at  2.44 was replaced by a sharp methylene singlet at  4.70, consistent with substitution with oxygen, while the neighboring methine proton of the isoxazole ring was modestly downfield shifted from  6.45 to  6.66. No other changes were apparent in the NMR spectrum.

Metabolite M4. The MS1 full scan spectrum of M4 showed the ion at m/z 571 as the most abundant ion, which represented a loss of 28 Da. Its product ion spectrum (Fig. 3d) showed that those fragment ions containing the P1/P1' moiety (e.g., m/z 199 and 427) lost 28 Da as well, consistent with the carboxylic acid metabolite of AG7088. This assignment was consistent with the LC-NMR data, which showed a loss of the ethyl protons and a downfield shift of the olefinic proton from  5.23 to  5.32. The identity of this metabolite was confirmed by comparison with the MS/MS and LC-NMR spectra of a synthetic standard of the acid metabolite (AG7185).

Metabolites M5 and M6. The MS1 full scan and product ion (Fig. 3e) mass spectra of M5 indicated that it is the carboxylic acid analog of M1/M2, and must therefore represent a pair of diastereomers. Finally, the MS1 full scan and product ion (Fig. 3f) mass spectra of M6 indicated that it is the carboxylic acid analog of M3. These compounds were not assessed by LC-NMR.

	Discussion

Top Abstract Introduction Results Discussion References

While the sensitivity of LC-MS/MS provides rapid metabolite information during drug discovery or development, it often lacks the accuracy of pinpointing the precise location of metabolism. LC-NMR, on the other hand, offers detailed structural information that is complementary to the LC-MS/MS data. In this case, the complementary results obtained from LC-MS/MS and stop-flow LC-NMR were critical in providing definitive structural information for these metabolites. The overall metabolic pathway for AG7088 as proposed from these data is summarized in Fig. 5. The LC-MS/MS data were sufficient to determine the structure of hydrolysis product M4, while LC-NMR provided confirmation of its identity. LC-MS/MS also provided a highly plausible structure for M3 and localized the site of oxidation for M1 and M2 to the P1 moiety, while LC-NMR allowed the specific sites of oxidation to be identified. Clearly, the combination of hyphenated techniques provided a powerful means of elucidating the structures of all of the metabolites with very limited quantities of material.

View larger version (24K): [in this window] [in a new window]   Fig. 5.   Metabolic pathway for AG7088 in vitro, as deduced from LC-MS/MS and LC-NMR analysis.

The metabolic pathway of AG7088 was evaluated across species, including mouse, rat, rabbit, monkey, dog, and human. Hydrolysis of the ethyl ester to produce M4 is the predominant pathway in all species, with rodents and rabbits being the most active, followed by monkeys. Hydrolysis occurred in dog and human liver microsomes at a more moderate rate. Hydroxylation on the methyl position of the methylisoxazole ring to give M3 also appeared to be a prominent pathway in monkey, dog, and human liver microsomes. Two additional metabolites, M1 and M2, were identified as a pair of diastereomers with a hydroxy group at the -position of the P1-lactam moiety. Rabbit, dog, and human liver microsomes formed more M1/M2 when compared with mouse, rat, and monkey liver microsomes. M1/M2 were subject to further hydrolysis to form secondary metabolite M5, which represents a putative mixture of diastereomers. The other hydroxylated metabolite, M3, was also hydrolyzed to its corresponding carboxylic acid, M6. Secondary metabolites M5 and M6 were more prominent products in monkey, rabbit, rat and mouse liver microsomes.

The metabolic profile of AG7088 in dog liver microsomes resembled that of human liver microsomes, indicating that the dog may be the most appropriate animal model relative to humans for exposure to AG7088 and its metabolites. Although AG7088 appeared to be stable from esterase hydrolysis studies conducted in dog and human plasma (Kosa et al., 1999), the experiments described herein demonstrated that extensive hydrolysis can take place in the liver, and may therefore result in significant first pass metabolism and poor oral bioavailability (Harr et al., 1999). This hypothesis was supported by in vivo studies in the dog, in which the bioavailability of AG7088 was only 8% (T. Tuntland and C. Lee, unpublished results). Collectively, these findings favored a clinical development program that localized the delivery of AG7088 to the nasal cavity, which is believed to be the primary loci of rhinoviruses. The major hydrolysis product of AG7088, M4, had significantly reduced binding affinity toward rhinovirus 3C protease (K_obs/I = 1.03 × 10⁴ M¹ s¹) and reduced antiviral activity (ED₅₀ = 12.3 mM) against HRV14 infected HI-HeLa cells.