Abstract
Two isomeric metabolites of GDC-0623 [5-((2-fluoro-4-iodophenyl)amino)-N-(2-hydroxyethoxy)imidazo[1,5-a]pyridine-6-carboxamide], a mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) kinase inhibitor, were identified in radiolabeled mass balance studies in rats and dogs (approximately 5% in excreta) and were also observed in human circulation (nonradiolabeled). Mass spectrometric data indicated that both metabolites had formed a new ring structure fused to the imidazopyridine core. Given their unusual structures, we conducted experiments to elucidate their chemical structures and understand the mechanisms for their formation. For the first metabolite, M14, a pyrazol-3-ol ring was generated by N-N bond formation between the aniline and hydroxamate. For the second metabolite, M13, an imidazol-2-one was generated by a Hofmann-type rearrangement that involved C-C bond cleavage and C-N bond formation. Both reactions were catalyzed by CYP2C9 and CYP2C19. M14 was generated directly from GDC-0623 and we speculate that its formation was via oxidative activation of the hydroxamic ester by cytochrome P450 (P450) and intramolecular nucleophilic displacement of the ester side chain. M13 (the rearranged metabolite) formed from the N-reduced hydroxamate (amide) and not from GDC-0623 directly. We propose for M13 that a P450-mediated reaction formed a cationic amide intermediate, which enabled the molecular rearrangement of the imidazopyridine core migrating from the amide carbon to the nitrogen and subsequent cyclization reaction. Each of these metabolic pathways constitutes a novel biotransformation mediated by P450 enzymes.
Introduction
GDC-0623 [5-((2-fluoro-4-iodophenyl)amino)-N-(2-hydroxyethoxy)imidazo[1,5-a]pyridine-6-carboxamide] is a potent and selective ATP-uncompetitive inhibitor of mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) kinase 1 (Robarge et al., 2014). Because of its promising antitumor activities, particularly against V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog KRAS-driven cancer cell lines (Hatzivassiliou et al., 2013), it was advanced to the clinic for treating patients with locally advanced or metastatic solid tumors.
The absorption, distribution, metabolism, and excretion properties of GDC-0623 were studied in vivo with radiolabeled studies in rats and dogs (unpublished data) and were assessed during the phase 1 dose escalation study in humans (El-Khoueiry et al., 2013). In the preclinical species, the excretion profiles were very similar. The majority of the dose (approximately 76%) was recovered in feces, with a high proportion delivered in bile (approximately 69% of the dose), and a minor component was recovered in urine (approximately 13% of the dose). Metabolism was the primary route for clearance, with glucuronidation as the predominant metabolic pathway. N-O reductive cleavage and hydrolysis of the hydroxamate and oxidation at the core or ester side chain were minor pathways. Metabolite profiling of clinical plasma samples suggested that, in general, the metabolic pathways for GDC-0623 in animals and humans were largely conserved.
The in vivo metabolic characterization for GDC-0623 revealed two isomeric metabolites named M13 and M14, plus a corresponding glucuronide conjugate. Although they were minor in abundance in rats and dogs (approximately 5% of the administered dose), they were of particular interest to us because of their unique structures and the novelty of the metabolic pathways that generated them. Furthermore, these metabolites were in plasma samples for all species, including humans. In this report, we describe the discovery of the two metabolites (M13 and M14), the structural characterization that defined their chemical structures, and the mechanistic experiments that revealed how they were generated by cytochrome P450 (P450)–catalyzed reactions.
Materials and Methods
Chemicals and Reagents
GDC-0623, d4-GDC-0623 (2-hydroxyethoxy-1,1,2,2-d4 analog), and M15 (amide corresponding to reduced GDC-0623) were synthesized at Genentech, Inc. (South San Francisco, CA). [14C]GDC-0623 ([Aniline-U-14C]GDC-0623) was synthesized at Selcia Limited (Essex, UK) with a specific activity of 50.63 mCi/mmol (radiochemical purity = 98.2%). 1-(2-Fluoro-4-iodophenyl)-1H-diimidazo[1,5-a:4′,5′-e]pyridin-2(3H)-one (corresponding to M13) was synthesized by Provence Technologies (Marseille, France) (Supplemental Fig. 1; Supplemental Methods). Human liver microsomes (pool of 150 donors, mixed sex) and P450 Supersomes were purchased from BD Biosciences (San Jose, CA). Deuterated solvents (d2-water and d3-acetonitrile (d, >99.9%)) were from Cambridge Isotope Laboratories (Andover, MA). All other reagents were of analytical grade or higher quality from commercial vendors.
Sample Preparation for Metabolite Profiling
In vivo samples were collected in radiolabeled [14C]GDC-0623 single oral-dose studies in rats (7 mg/kg; 375 μCi/kg) and dogs (1 mg/kg; 20 μCi/kg) (Supplemental Methods). An equal percentage of each excreta collection was pooled to create a single sample for each matrix/animal that contained >95% of the radioactivity eliminated by the excretion route. Plasma was pooled proportional to time to assess the exposures of GDC-0623 and its metabolites (Hop et al., 1998). For clinical samples, plasma was for day 14 (0–6 hours postdose) from patients on a regimen of daily oral administration at 65 mg. The analysis of metabolites was deemed investigative and used plasma remaining after bioanalysis for parent pharmacokinetics.
Plasma and fecal samples were repeatedly extracted with acetonitrile (ACN) via sonication, vortex mixing, and then centrifugation at 16,000×g for 5 minutes until the full sample radioactivity was recovered or no additional radioactivity was found in the extracts. The pooled extracts were evaporated to near dryness under nitrogen at 35°C and then redissolved with an appropriate volume of solvent for injection to the liquid chromatography (LC) column. Urine and bile samples were either diluted with solvent or concentrated under a vacuum until they were an appropriate concentration of radioactivity for profiling; samples were then centrifuged to remove solids and the supernatants were injected onto the LC column.
LC–Tandem Mass Spectrometry and Radioprofiling of Metabolites
The Accela LC system consisted of a solvent pump with a built-in degasser, temperature-controlled autosampler, and photodiode array detector (Thermo Scientific, San Jose, CA). Chromatography was performed on a Luna C18 column (3 µm, 4.6×150 mm; Phenomenex, Torrance, CA) with mobile phases A [10 mM ammonium acetate in water with 0.01% formic acid, pH 5.0] and B [10 mM ammonium acetate in ACN/water (90/10) with 0.01% formic acid]. The flow rate was 1.0 ml/min. The high-performance liquid chromatography (HPLC) gradient was as follows: 10% B held for 5 minutes, increased to 35% at 50 minutes, to 50% at 58 minutes, and to 95% at 71 minutes and held for 5 minutes before returning to 10% at 76.1 minutes for 4 minutes. The column eluent was split postcolumn with three parts directed to a βRAM5c online radiomatic detector (IN/US, Tampa, FL) and one part directed to the mass spectrometer. For radiometric detection, the column flow was mixed with InFlow scintillation fluid (IN/US) at a 2:1 ratio and active counting mode was enabled. The flow cell volume was 500 µl. For this study, we used Laura (IN/US) and Xcalibur (Thermo Scientific) acquisition and processing software for radiometric data and LC–tandem mass spectrometry (MS/MS) data, respectively.
Human plasma and in vitro samples were profiled using a shorter, higher-pressure LC method. To confirm that the same metabolite assignments were made, selected rat or dog samples were analyzed using this method and MS/MS spectra were compared. The column was a Hypersil Gold C18 (1.9 μm, 2.1 × 100 mm; Thermo Scientific) and mobile phase A comprised 0.1% formic acid in water, whereas mobile phase B comprised 0.1% formic acid in ACN. The flow rate was 0.4 ml/min. The HPLC gradient was as follows: 10% B held for 6 minutes, increased to 20% at 16 minutes, to 45% at 19 minutes, and to 98% at 20 minutes and held for 3 minutes; the column was then reequilibrated at 10% B for the next injection.
MS detection of metabolites was performed with an LTQ-Orbitrap mass spectrometer equipped with an electrospray ion source (Thermo Scientific). The ion source voltage was 4.5 kV. The heated capillary temperature was 350°C. Auxiliary and sheath gases were set to 10 and 25 units, respectively. The scan-event cycle consisted of a full mass scan at a resolving power of 30,000 [at a mass-to-charge ratio (m/z) of 400] and data-dependent product ion scans with collision-induced dissociation or higher-energy collision-induced dissociation at a resolving power of 7500. Accurate mass measurements were performed using an external calibration.
Hydrogen/Deuterium Exchange.
The HPLC mobile phases were replaced with deuterated solvents [A: 10 mM ammonium acetate in D2O with 0.01% formic acid; and B: 10 mM ammonium acetate in d3-ACN/D2O (90/10)] and the MS data for protonated/deuteronated molecular and product ions were used to identify the number of exchangeable protons.
Structure Determination M13 and M14 by Nuclear Magnetic Resonance
Synthesized GDC-0623 (approximately 5 mg), synthesized M13 (approximately 1.5 mg), and isolated M14 (approximately 0.2 mg) were separately solubilized in CD3OD (99.8% D + 0.05% v/v trimethylsilane; Cambridge Isotope Laboratories). Details of M13 synthesis are presented in the Supplemental Methods. Nuclear magnetic resonance (NMR) measurements were performed on an Avance 3 600-MHz spectrometer equipped with a 1.7-mm TCI or 5-mm TXI Cryoprobe (Bruker Daltronics, Billerica, MA). The sample temperature was maintained at 28°C for all data collections and spectra were acquired using standard pulse sequences.
In Vitro Investigations
P450 Phenotyping.
P450 Supersomes (CYP1A1, CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, and CYP3A5) were incubated at a constant amount of P450 (40 pmol/ml) supplemented with NADPH (1 mM) in potassium phosphate buffer (100 mM, pH 7.4) with MgCl2 (3 mM) and GDC-0623 or M15 (1 µM) for 60 minutes at 37°C. At the end of the incubation, the reactions were quenched with an equal volume of cold ACN. The supernatants were diluted with an equal volume of water prior to LC-MS/MS analysis.
Kinetic Deuterium Isotope Effect for M14 Formation.
Deuterated (d4) and nondeuterated (d0) GDC-0623 were separately incubated with CYP2C19 Supersomes to determine the kinetics of M14 formation. The incubation samples (125 µl) contained recombinant P450 (4 pmol/ml), GDC-0623 (0.1–10 µM), and NADPH (1 mM) in potassium phosphate buffer (100 mM, pH 7.4) with MgCl2 (3 mM). Incubations lasted 10 minutes at 37°C with shaking and were stopped with 1 volume of cold ACN, which contained the internal standard. The supernatants were diluted with an equal volume of water prior to LC-MS/MS analysis. M14 was monitored using the transition m/z 394.9 → 268.0. The plots of reaction rates versus substrate concentrations were fit by nonlinear regression with Michaelis–Menten kinetics using Prism software (version 6.05; GraphPad, San Diego, CA) to obtain parameter estimates for Km and Vmax. Ratios of the observed Vmax and intrinsic clearance (determined as Vmax/Km) for the reactions with d0- and d4-GDC-0623 were calculated to assess the kinetic deuterium isotope effect (KDIE).
Results and Discussion
In elucidating the metabolites of GDC-0623 from mass balance studies in rats and dogs, M13 and M14 were identified. These metabolites were isomeric and, based on their molecular ions, formed by N-O reductive cleavage and loss of two hydrogens (or reactions that result in the same net change in molecular formula), which was consistent with cyclization reactions. Formation of new fused ring structures by metabolism are relatively uncommon (Kalgutkar et al., 2005), and M13 and M14 were even more curious because a limited number of diverse cyclized metabolites could be rationalized. This prompted our investigations to elucidate their chemical structures and mechanisms to describe their formation.
M13 and M14 were chromatographically separated analytes with identical protonated molecular ions [m/z 394.9798, C14H9FIN4O (−0.5 ppm)] and product ion spectra (Fig. 1). The base peak in MS/MS spectra (collision-induced dissociation of MH+ ions) was at m/z 368, formed by the neutral loss of hydrogen cyanide (HCN) from the fused imidazole. Although HCN loss was observed in data-dependent product ion scan experiments for GDC-0623 and M15, cleavages of the hydroxamate or amide moieties were predominant. A more facile loss of HCN for M13 and M14 provided some evidence of cyclization at the hydroxamate. Losses of iodine radical (−127 Da) from the molecular ion and the m/z 368 ion generated ions at m/z 268 and 241, respectively. A neutral loss of fluoro-iodobenzene from the ion at m/z 368 provided the ion at m/z 148. An ion corresponding to fluoro-iodoaniline was observed at m/z 248 and showed iodine radical loss to give the ion at m/z 121. The hydrogen/deuterium exchange experiment supported cyclization reactions since M13 and M14 had molecular ion mass shifts of 2 Da, whereas M15 and GDC-0623 had mass shifts of 4 Da (Table 1). This was also consistent with the aromaticity of the parent molecule, which could not lose two exchangeable hydrogen atoms by desaturation.
MS/MS spectra (collision-induced dissociation of MH+ ions) and proposed product ions for GDC-0623 and its metabolites under study.
Observed protonated and deuteronated molecular ion and product ions for GDC-0623 and its metabolites in hydrogen/deuterium exchange experiments
The structure for M14 was determined by generating it in vitro using recombinant P450, isolating it in pure form by preparative HPLC, and analyzing the isolate by NMR. The pyrazol-3-ol ring was generated by N-N bond formation between the aniline and hydroxamate. Two-dimensional NMR data (gradient-selected COSY, gradient Heteronuclear Single Quantum Coherence, and gradient Hetereonuclear Multiple Bond Coherence) and 1H J values established identical connectivity for GDC-0623 and M14 at the fused imidazole and fluoroiodobenzene. M14 cyclization to pyrazol-3-ol at the aniline was inferred from 13C and 1H chemical shift differences between GDC-0623 and M14 (Supplemental Table 1). Because of the limited amount of M14 isolated, 15N data were not available to discreetly discern between the two tautomeric forms, pyrazol-3-ol or pyrazol-3-one.
The chemical structure for M13 was confirmed by matching it with a synthetic standard. Standard and test samples containing M13 were mixed and chromatographed using two different column chemistries (C18 and pentafluorophenyl) and there was no separation of the M13 peak from the standard peak or non-Gaussian peak shape. MS/MS spectra for the standard and M13 matched. The synthetic scheme to provide the standard (i.e., installation of the carbonyl carbon as the last step) and characterization data unambiguously defined the embedded urea structure for M13 (Supplemental Table 1), which indicated that a molecular rearrangement had occurred to form the imidazol-2-one structure.
In rats and dogs, the cyclized metabolites were observed mainly as a glucuronide conjugate, M2, in bile (5.1% of the administered dose in rats and 1.0% of the dose in dogs). The correspondence of M2 as the conjugate of M14 was confirmed by its MS/MS spectra and β-glucuronidase treatment. M2 was also in urine (2.1% of the dose in rats and trace levels in dogs). M2 in bile was presumably secreted to the gastrointestinal tract, cleaved during transit by microflora enzymes, and eliminated as M14 in feces (2.7% of the dose in rats and 4.6% of the dose in dogs). M13 was at lower levels than M2 or M14 (≤1% of the dose in rats and dogs in urine). M2, M13, and M14 were detected as circulating metabolites in rats and dogs but were below the limits of measurement attained by radiodetection. In human plasma samples (steady state), M2 and M14 were measured with MS responses greater than 10% of the unchanged parent, and M13 was detected at lower abundance. Overall, there was good consistency across species that the cyclization reactions had occurred as minor pathways in the full metabolism for GDC-0623 (Supplemental Table 2). A related metabolite was M15, the amide resulting from N-O reductive cleavage of GDC-0623, which was found in fecal samples and identified by matching the chemical standard. M15 accounted for 4.0% and 14.8% of the dose in rats and dogs, respectively (Supplemental Table 2).
With the structures for the metabolites confirmed, we sought to understand the metabolic pathways and how they diverged to form two distinct ring structures. Incubations of GDC-0623 with human liver microsomes formed M14 and, to a much lesser extent, M13. The formation of the metabolites was time and NADPH dependent, which was consistent with P450-mediated reactions. To investigate the enzymology, GDC-0623 and M15 were separately incubated with recombinant P450s. Full-scan analysis of incubates with CYP2C enzymes showed M14 and M13 to be the predominant metabolites formed from GDC-0623 and M15, respectively. This indicated that generating M13 from GDC-0623 required at least two metabolic steps: first, N-O reductive cleavage to provide M15; and then, transformation of M15 to M13. For the reactions of GDC-0623 to M14 and M15 to M13, CYP2C19 was the most efficient enzyme. CYP2C9 formed the metabolites to a lesser extent and the other tested P450 isoforms did not form them. We conducted exploratory experiments to investigate the reductive conversion of GDC-0623 to M15 but have not yet identified the enzyme responsible (unpublished data). Incubations of GDC-0623 with liver cytosolic fractions to probe aldehyde oxidase-mediated reduction did not generate M15. This differentiated this reaction from hydroxamic acid to amide in CP-544439 (4-[4-(4-fluorophenoxy)-benzenesulfonylamino]tetrahydropyran-4-carboxylic acid hydroxyamide) by aldehyde oxidase (Obach, 2004).
We postulate that the pyrazol-3-ol structure for M14 is formed by nucleophilic displacement by the aniline nitrogen at the hydroxamate (pathway A, Fig. 2). This reaction requires activating the hydroxamate to become an electrophilic reaction site and we speculate that the initiating step is P450 oxidation at the 2-hydroxyethoxy side chain (step A1), which transforms it to a good leaving group (2-hydroxyacetaldehyde in step A2). The activated intermediate also partitioned via an O-dealkylation reaction to the hydroxamic acid derivative of GDC-0623 (step A2b). The detected amounts of hydroxamic acid correlated with those of M14 in vitro, with the most in incubations of GDC-0623 with recombinant CYP2C19 and less with recombinant CYP2C9. The hydroxamic acid could be a postulated precursor of M14, although water would be a poor leaving group for the cyclization reaction. Since we had a fully deuterated ethoxy analog (d4-GDC-0623 synthesized for bioanalytic purposes), we tested whether a KDIE was discerned for forming M14 using recombinant CYP2C19. Although the reaction rates were consistently lower with d4-GDC-0623, there was no significant observed KDIE, with Vmax(kH/kD)obs of 1.11 and intrinsic clearance (kH/kD)obs of 1.04 (Supplemental Fig. 2). We presume that the isotope effect for C-H bond breaking in GDC-0623, which enabled forming M14, was masked by other rate-limiting steps of the enzymatic cycle (Nelson and Trager, 2003). It is also possible that hydroxamic acid esters react via mechanisms that are distinct from O-dealkylation reactions and this is the focus of future investigative experiments.
Proposed mechanisms for the formation of M14 and M13 from GDC-0623 and M15.
A distinct mechanism was needed to rationalize the reaction that forms M13 (pathway B, Fig. 2). This reaction needs to be consistent with M13 forming from M15 and not directly from GDC-0623, and it must provide a basis to break a C-C bond and form a new C-N bond. We propose as the initiating step, a two-electron oxidation by P450 at fluoroiodoaniline (step B1) to form an iminoiodonium intermediate (IB1). This reaction is similar to that previously described for N-acetyl-p-benzoquinone imine formation from acetaminophen (Nelson et al., 1981) and for ipso addition of glutathione and dehalogenation of dihalogenated anilines (Zhang et al., 2011). The imine can intramolecularly accept a hydride from the amide to provide an unstable cationic form of M15 that is primed for migration of imidazopyridine from the carbonyl carbon to nitrogen (step B2). The resulting formamide intermediate (IB2) can undergo a ring-forming reaction with the aniline nitrogen adding at the carbonyl carbon to give a fused 1,3-dihydro-2H-imidazol-2-one (M13). The migration reaction has analogies to reactions for anionic intermediates in the Lossen rearrangement and Hofmann degradation. In the Lossen rearrangement, an anionic species forms from a hydroxamic acid O-derivative (acyl, sulfonyl, or phosphoryl) by hydrogen abstraction by a base (Yale, 1943). In the Hofmann degradation, a primary amide reacts with bromine to form an N-bromoamide, which is converted to its anionic form by proton abstraction (Wallis and Lane, 2011). It is notable that a Lossen rearrangement was proposed for forming sulfonamide and sulfonic acid metabolites of CP-544439 (Dalvie et al., 2008). The current case of GDC-0623 offers direct evidence for this C-C bond–breaking and C-N bond–forming rearrangement reaction because no parts of the molecule are lost. It is also, to our knowledge, the clearest demonstration to date that this rearrangement can result from P450 activities.
The diversity of P450s in protein structure and function allow them to act on a large array of substrates for, most commonly, oxidation reactions. In some instances, the products of P450 reactions have electronic structures that are inherently unstable and allow them to undergo rearrangement reactions. Examples have been well described in previous reviews (Ortiz de Montellano and Nelson, 2011; Guengerich and Munro, 2013). Here, we report two unusual ring-forming reactions that were catalyzed by P450s in the metabolism of GDC-0623. The first of these reactions forms a pyrazole structure by forming an intramolecular N-N bond. The other reaction was a Hofmann-like rearrangement that we rationalize to result from forming a cationic amide intermediate that is stabilized by C-C bond breaking and C-N bond forming. This case study reemphasizes the variety of novel reactions that are catalyzed by P450 enzymes.
Acknowledgments
The authors thank their colleagues for valued collaborations for this work: Dr. Luna Musib (Clinical Pharmacology) for design and conduct of clinical studies for GDC-0623, Dr. Alan Deese (Small Molecule Pharmaceutical Sciences) for the preliminary NMR analysis of M14, and Drs. Jiao Jiao, Mark Zak, and Bing-Yan Zhu (Discovery Chemistry) for scientific discussions of the reactions and providing the synthetic standard of M13. They also thank Drs. Babak Sayah and Julien Hugon (Provence Technologies) for chemical synthesis of M13, as well as attendees of the 2015 Gordon Research Conference on Drug Metabolism for spirited scientific discussions of these metabolites.
Authorship Contributions
Participated in research design: Takahashi, Ma, Robinson, Yue, Choo, Khojasteh.
Conducted experiments: Takahashi, Robinson.
Contributed new reagents or analytic tools: Robinson.
Performed data analysis: Takahashi, Ma, Robinson, Khojasteh.
Wrote or contributed to the writing of the manuscript: Takahashi, Ma, Robinson, Khojasteh.
Footnotes
- Received September 14, 2015.
- Accepted October 2, 2015.
↵1 Current affiliation: Novartis Institutes for BioMedical Research, Emeryville, California.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- ACN
- acetonitrile
- CP-544439
- 4-[4-(4-fluorophenoxy)-benzenesulfonylamino]tetrahydropyran-4-carboxylic acid hydroxyamide
- GDC-0623
- 5-((2-fluoro-4-iodophenyl)amino)-N-(2-hydroxyethoxy)imidazo[1,5-a]pyridine-6-carboxamide
- HCN
- hydrogen cyanide
- HPLC
- high-performance liquid chromatography
- KDIE
- kinetic deuterium isotope effect
- LC
- liquid chromatography
- m/z
- mass-to-charge ratio
- MS
- mass spectrometry
- MS/MS
- tandem mass spectrometry
- NMR
- nuclear magnetic resonance
- P450
- cytochrome P450
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics
References
In this issue
Unusual Biotransformations of GDC-0623 by P450
