Abstract
During a medicinal chemistry campaign to identify inhibitors of the hepatitis C virus nonstructural protein 5B (RNA-dependent RNA polymerase), a bicyclo[1.1.1]pentane was introduced into the chemical scaffold to improve metabolic stability. The inhibitors bearing this feature, compound 1 [5-(3-(bicyclo[1.1.1]pentan-1-ylcarbamoyl)-4-fluorophenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide] and compound 2 [5-(3-(bicyclo[1.1.1]pentan-1-ylcarbamoyl)phenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide], exhibited low turnover in incubations with liver S9 or hepatocytes (rat, human), with hydroxylation of the bicyclic moiety being the only metabolic pathway observed. In subsequent disposition studies using bile duct–cannulated rats, the metabolite profiles of bile samples revealed, in addition to multiple products of bicyclopentane oxidation, unexpected metabolites characterized by molecular masses that were 181 Da greater than those of compound 1 or 2. Further liquid chromatography/multiple-stage mass spectrometry and nuclear magnetic resonance analysis of the isolated metabolite of compound 1 demonstrated the presence of a phosphocholine (POPC) moiety bound to the methine carbon of the bicyclic moiety through an ester bond. The POPC conjugate of the nonstructural protein 5B inhibitors was assumed to result from two sequential reactions: hydroxylation of the bicyclic methine to a tertiary alcohol and addition of POPC by cytidine-diphosphocholine:1,2-diacylglycerol cholinephosphotransferase, an enzyme responsible for the final step in the biosynthesis of phosphatidylcholine. However, this pathway could not be recapitulated using cytidine-diphosphocholine–supplemented liver S9 or hepatocytes because of inadequate formation of the hydroxylation product in vitro. The observation of this unexpected pathway prompted concerns about the possibility that compounds 1 and 2 might interfere with routine phospholipid synthesis. These results demonstrate the participation in xenobiotic metabolism of a process whose function is ordinarily limited to the synthesis of endogenous compounds.
Introduction
After a decade of research and clinical trials, the treatment of hepatitis C virus (HCV) infection, which plagues nearly 3% of the world’s population and leads potentially to progressive chronic liver diseases, has shifted away from the use of pegylated interferon and ribavirin—medicines that offer only suboptimal sustained virological responses and produce severe side effects (Eltahla et al., 2015). Instead, direct-acting antiviral agents have emerged as the current standard of care to treat HCV infection. These drugs are directed against multiple nonstructural (NS) proteins that are essential to the virus life cycle, including the NS5B RNA-dependent RNA polymerase, the NS5A replication complex, and the NS3/4A protease (Lindenbach and Rice, 2005). Structural and functional characterization of NS5B revealed a right-hand–like motif composed of one active site and multiple allosteric sites, including two palm sites, two finger sites, and one thumb site (Ago et al., 1999; Bressanelli et al., 1999; Lesburg et al., 1999). Identification of these structural features laid a foundation for the design of potent antiviral agents that, by binding to these NS5B allosteric sites, could obstruct conformational changes of the polymerase that are required for initiation and elongation of RNA strands (Caillet-Saguy et al., 2011; Eltahla et al., 2014). Thus far, anti-HCV drug discovery targeting NS5B has identified both non-nucleotide inhibitors (e.g., BMS-791325 [(5aR)-12-cyclohexyl-N-(N,N-dimethylsulfamoyl)-3-methoxy-5a-((1R,5S)-3-methyl-3,8-diazabicyclo[3.2.1]octane-8-carbonyl)-4b,5,5a,6-tetrahydrobenzo[3,4]cyclopropa[5,6]azepino[1,2-a]indole-9-carboxamide], dasabuvir) and nucleotide inhibitors (e.g., sofosbuvir).
Recently, potent non-nucleotide NS5B inhibitors targeting the palm site of the HCV RNA-dependent RNA polymerase were evaluated alongside inhibitors of other key HCV enzymes, including NS5A, NS3/4A, and NS5B (the thumb site), for possible use in combination therapy (Gao et al., 2010; Gentles et al., 2014; Scola et al., 2014). During optimization to improve the metabolic stability of these NS5B inhibitors featuring a 6-substituted furo[2,3-b]pyridine core, multiple isosteric replacements of a t-butyl moiety were explored including a bicyclo[1.1.1.]pentane (Fig. 1). As a precedent for this approach, a bicyclo[1.1.1]pentane moiety was used during the discovery of metabotropic glutamate receptor-1 antagonists as a phenyl replacement that maintained flanking pharmacophores in a coplanar orientation (Pellicciari et al., 1996; Costantino et al., 2001). In another example involving the optimization of a group of γ-secretase inhibitors, the use of bicyclo[1.1.1]pentane as a phenyl isostere imparted improvements in metabolic stability, aqueous solubility, and permeability that were attributed to changes in physicochemical properties (Stepan et al., 2012).
Structures of two selected HCV NS5B inhibitors featuring a bicyclo[1.1.1]pentane.
Compared with NS5B inhibitors featuring t-butyl groups in the same position, the bicyclo[1.1.1.]pentane analogs evaluated in this study retained comparable antiviral activity against genotype 1a, 1b, and 2a proteins (<10 nM) and also avoided rapid hydroxylation by cytochrome P450 (P450) enzymes. Accordingly, they were advanced for further evaluation of absorption, distribution, metabolism, and excretion properties, including the characterization of metabolite profiles in liver S9, hepatocytes, and bile duct–cannulated (BDC) rats. After the detection of unexpected metabolites in rat bile, multiple approaches including liquid chromatography (LC)/multiple-stage mass spectrometry (MSn) and nuclear magnetic resonance (NMR) were employed to elucidate the structures of these metabolites with the aim to foster a better understanding of their mechanism of formation.
Materials and Methods
Chemicals and Liver Subcellular Fractions.
NADPH, potassium phosphate (monobasic and dibasic), cytidine diphosphocholine (CDP-choline), ammonium formate, EDTA, formic acid, and deuterated methanol (methanol-d4) were purchased from Sigma-Aldrich (St. Louis, MO). Magnesium chloride (1 M) solution was obtained from ThermoFisher Scientific (Grand Island, NY). High-pressure liquid chromatography (HPLC)–grade water and acetonitrile were obtained from Mallinckrodt Baker (Phillipsburg, NJ). Human liver S9 (pooled from 20 male and female donors) and Aroclor-induced rat liver S9 (pooled from male animals) were obtained from BD Biosciences (Bedford, MA) and from Xenotech (Lenexa, KS), respectively. Cryopreserved human hepatocytes (single male donor) and rat hepatocytes (from male animals) were provided by CellzDirect, Inc (Tucson, AZ). Everolimus was purchased from Cell Signaling (Danvers, MA). Compound 1 [5-(3-(Bicyclo[1.1.1]pentan-1-ylcarbamoyl)-4-fluorophenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide] and compound 2 [5-(3-(bicyclo[1.1.1]pentan-1-ylcarbamoyl)phenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide] were synthesized and characterized by Bristol-Myers Squibb (Wallingford, CT) (Fig. 1; refer to Yeung et al., 2014).
In Vitro Metabolism.
Compounds 1 and 2 (10 µM each) were studied in incubations (0.5 ml) with either human liver S9 (2 mg/ml) or Aroclor-induced rat liver S9 (2 mg/ml) in potassium phosphate buffer (100 mM, pH 7.4). The reactions (n = 2) were initiated by addition of NADPH (1 mM) and carried out at 37°C. At 0 and 30 minutes, aliquots (150 µl) of the reaction mixtures were collected, and reactions were terminated by adding 3 volumes of acetonitrile. Proteins were then removed using a Strata filter plate (Phenomenex, Torrance, CA) by centrifugation (5804 R; Eppendorf, Hamburg, Germany) at 2000g for 3 minutes. The filtrate was collected in a 96-well plate and evaporated to 25% of the original volume under N2 gas. The remaining aqueous phase was analyzed using HPLC/UV/MS.
To verify the presence of active CDP-choline:1,2-diacylglycerol cholinephosphotransferase (CPT) in vitro, everolimus (10 µM) was incubated with human liver S9 or Aroclor-induced rat liver S9 supplemented with CDP-choline (1 mM), magnesium chloride (5 mM), and EDTA (10 mM) in a Tris-HCl buffer (100 mM, pH 8.5) (Wilgram et al., 1960; Arthur and Choy, 1984). The reactions were carried out at 37°C for 120 minutes, and the samples were then processed as described above. In a separate experiment, compound 1 (10 µM) was incubated under the same experimental conditions in vitro, but with an additional cofactor, NADPH (1 mM), added to facilitate phase I reactions.
The in vitro metabolism of compounds 1 and 2 was also examined in incubations with cryopreserved primary human or rat hepatocytes. The frozen cells (one tube each) were thawed in a 37°C water bath and then added to 3 ml cryopreserved hepatocyte recovery medium (Life Technologies, Grand Island, NY) for human cells or to 3 ml Williams’ medium E supplemented with a hepatocyte maintenance supplement pack (Life Technologies) for rat cells. After a brief centrifugation at 100g for 10 minutes (for human hepatocytes) or at 55g for 3 minutes (for rat hepatocytes), the cell pellet was resuspended in 0.5 ml prewarmed hepatocyte incubation medium (Life Technologies), and cell viability and yield were determined using trypan blue dye exclusion staining. The hepatocytes (with at least 80% viability) in suspension (1 × 106 cells/ml) were incubated with a substrate (10 μM) for 0 and 2 hours at 37°C in a humidified CO2 (5%) incubator. The reaction was stopped by addition of an equal volume of acetonitrile. After centrifugation at 1500g for 20 minutes, the supernatant was evaporated to 50% of the original volume using N2 gas, and the remaining aqueous phase was analyzed using HPLC/UV/MS.
The activity of CPT was also assessed by incubating everolimus (10 µM) with rat hepatocytes (1 × 106 cells/ml) under the same conditions described above and by analyzing for the presence of the phosphocholine (POPC) conjugate reported previously (Zollinger et al., 2008).
BDC Rat Study.
All animal procedures were reviewed and approved by the Animal Care and Use Committee at Bristol-Myers Squibb Co. Compounds 1 and 2 were formulated (0.67 mg/ml) in a vehicle of dimethyl acetamide/PEG400 [10:90 (v/v)] and administered intravenously (2 mg/kg body weight; n = 3 for each compound) to BDC male rats (Hilltop Laboratory Animals, Inc., Scottdale, PA), and bile, plasma, urine, and feces were collected over 24 hours. Serial blood samples were collected from the jugular vein into K3EDTA-containing tubes at 0, 0.5, 1, 3, 6, and 24 hours postdose. Plasma samples were obtained by centrifugation at 4°C (1500–2000g) and stored at −80°C until analysis.
Aliquots (100 μl) of plasma samples were pooled across time points and mixed with 3 volumes of acetonitrile by vortexing. After precipitation of the protein pellet by centrifugation at 15,000g for 10 minutes, the supernatant was evaporated to 25% of the original volume under N2 gas. The remaining aqueous phase was analyzed by HPLC/UV/MS.
Aliquots (250 µl) of pooled bile or urine samples were mixed with 3 volumes of acetonitrile, centrifuged at 15,000g for 5 minutes, and the supernatants were evaporated to 25% of the original volume under N2 gas. The remaining aqueous phase was analyzed by HPLC/UV/MS.
Feces samples (2 g) were homogenized in 3 volumes of a potassium phosphate buffer (50 mM, pH 7.4), and extracted twice with ethyl acetate. After centrifugation at 15,000g for 5 minutes, the supernatants were dried using N2 gas. The extracts were resuspended in 300 µl solvent [water/acetonitrile/formic acid, 50:50:0.1 (v/v/v)] for HPLC/UV/MS analysis.
Isolation of 1-3.
Aliquots (70 ml) of bile from the BDC rats administered with compound 1 were extracted with ethyl acetate followed by n-butanol. The butanol extract was evaporated to dryness, redissolved in methanol/water [65:35 (v/v)], and extracted with chloroform that had been pre-equilibrated with methanol/water [65:35 (v/v)]. The aqueous methanol extract was further enriched by solid-phase extraction using a Waters Oasis HLB cartridge (1 g, 20 cc; Waters Corporation, Milford, MA). After sample application (10 ml in water), samples were eluted using a step-gradient of water/acetonitrile [9:1 (v/v)], followed by 3:1, 1:1, and 1:3 water/acetonitrile and 100% acetonitrile (10 ml each). 1-3 was detected primarily in the 1:1 water/acetonitrile eluent and was purified further on a Waters Sunfire C18 column (5 µm, 4.6 × 150 mm) using an Agilent 1100 series HPLC system (Agilent Technologies, Santa Clara, CA). The mobile phase consisted of solvent A (10 mM ammonium acetate in 5% acetonitrile) and solvent B (acetonitrile) with a linear gradient from 15% to 100% B over 25 minutes at a flow rate of 1.2 ml/min. The fractions were collected into Beckman 96-deep-well polypropylene blocks (Beckman Coulter Inc., Brea, CA) using an Agilent G1364C fraction collector. Fractions containing 1-3 were pooled and repurified to remove UV-transparent bile-related impurities using a shallow linear gradient on the same C18 column: 15%–75% solvent B over 25 minutes. The fractions containing 1-3 (approximately 25 µg) were evaporated to dryness under nitrogen at room temperature.
HPLC/UV/MS Analysis.
The HPLC/UV/MS system consisted of a Waters Acquity binary solvent manager, a Waters Acquity sample manager, a Waters Acquity photodiode array detector, and a Waters Xevo quadrupole time-of-flight mass spectrometer. Chromatographic separations were carried out on a Waters BEH C18 column (1.7 μm, 2.1 × 100 mm). The mobile phase consisted of water/acetonitrile/formic acid [95:5:0.1 (v/v/v)] (solvent A) and acetonitrile with 0.1% formic acid (solvent B) at a flow rate of 0.5 ml/min with a linear gradient as follows: 5% B isocratic for 0.5 minutes, 2%–35% B over 5.5 minutes, and finally, 35%–100% B over 3.5 minutes. The gradient was maintained for 1 minute and then returned to initial conditions for a 1.5-minute equilibration period. The eluent from the column was analyzed by the photodiode array detector (scanning 200–400 nm at 5 Hz) followed by the mass spectrometer equipped with an electrospray ionization source and operated in a positive-ion mode. To obtain maximum sensitivity based on the ionization of the parent molecules, the source temperature was set to 125°C, the desolvation temperature was 250°C, the capillary voltage was 3 kV, the sampling cone was 40 (arbitrary units), and the extraction cone was 1.7 (arbitrary units). All MS data were acquired using a low collision voltage (6 V) and a high-voltage ramp (25–50 V), and data were processed using Metabolynx software (Waters Corp.). Metabolite structures were assigned based on their LC/tandem mass spectrometry (MS/MS) product-ion mass spectra.
Everolimus and its conjugate were analyzed using the same LC/UV/MS system, except that the mobile phase was composed of solvent A (5 mM ammonium formate with 0.1% formic acid) and solvent B (methanol with 0.1% formic acid) (Bruns et al., 2015) and delivered at 0.3 ml/min using a linear gradient as follows: 30% B for 0.5 minutes, and then 30%–100% B over 5.5 minutes. The column was then re-equilibrated to the initial HPLC conditions for 2 minutes.
The isolated 1-3 was also analyzed using an LTQ Orbitrap mass spectrometer (Thermo Fisher, Waltham, MA) equipped with a nanoelectrospray source and operated in a positive-ion mode. The source voltage, capillary voltage, and tube lens voltage were set to 1.4 kV, 25 V, and 170 V, respectively, and the capillary temperature was set to 150°C. The scan ranges had a mass-to-charge ratio (m/z) 100–1000, 50–200, and 50–150 for MS/MS, MS3, and MS4 spectral data acquisition, respectively, with a normalized collision energy of 35% and an isolation width of m/z 2.0. Chromatographic separation was achieved using a Waters Nanoacquity system with a Waters Symmetry C18 column (5 µm, 20 × 180 mm) as a trap column connected to a Michrom Magic C18AQ (5 µm, 0.1 × 150 mm) column (Bruker-Michrom Inc., Auburn, CA) as an analytical column. After sample injection (3 µl) to the trap column and a 1-minute wash using solvent A [water/acetonitrile/formic acid, 95:5:0.1 (v/v/v)] at a flow rate of 10 µl/min, the trapped analyte was back-eluted with solvent B (acetonitrile) and further separated on the analytical column by applying a 10-minute linear gradient from 5% to 95% solvent B at a flow rate of 500 nl/min.
Relative abundances of drug-related components detected in the incubations were estimated according to their UV peak areas at the maximum absorption wavelength (λ = 303 nm) of the corresponding parent compounds. Molar extinction coefficients of compounds and their metabolites were assumed to be similar.
NMR.
The isolated 1-3 was reconstituted in 50 μl methanol-d4 and transferred to a 1.7-mm NMR tube that was placed in a 5-mm NMR carrier tube. All 1H spectra were acquired on a Bruker Avance 500 MHz spectrometer (Bruker BioSpin Corporation, Billerica, MA) equipped with a 5-mm TCI cryo probe, and 13C spectra were acquired at 125 MHz on the same instrument. The chemical shifts were referenced to the residual solvent (methanol-d4) signals of 4.78 and 49.3 ppm for 1H and 13C, respectively. The two dimensional spectra, including heteronuclear multiple quantum coherence (HMQC), heteronuclear multiple bond correlation (HMBC), and homonuclear correlation spectroscopy (COSY) experiments, were acquired according to standard protocols provided by the instrument manufacturer.
Results
In Vitro Metabolism.
The biotransformation of compounds 1 and 2 was studied using liver S9, primary hepatocytes, and BDC rats, and the results of these experiments are summarized in Table 1. The relative abundances of metabolites in vitro were expressed as percentages of the initial amount of the corresponding parent compound, based on UV peak areas. Relative amounts of metabolites in vivo were expressed as percentages of the total drug-related material detected in the BDC rat bile based on UV peak areas.
Summary of the metabolite profiles of compounds 1 and 2 obtained from incubations with liver S9 (human, rat), hepatocytes (human, rat), and bile (BDC rats)
The relative abundances (percentages) of the drug-related materials in incubations or matrices were estimated based on the UV peak areas (λ = 303 nm) assuming absorption coefficients of the parent compounds and their metabolites were similar. Metabolite names 1-1, 1-2, 2-1, and 2-2 are given in parentheses.
In incubations (30 minutes) of compound 1 with NADPH-supplemented liver S9 (human, Aroclor-induced rat) or in incubations (2 hours) with hepatocytes (human, rat), the only metabolites detected, 1-1 and 1-2, were identified as hydroxylation products based on their observed molecular weight shifts (16 Da) relative to the parent mass. Metabolite 1-1 was detected in both rat and human liver homogenates, whereas 1-2 was detected only in incubations with human liver S9. Both metabolites were minor components relative to the parent compound. In addition, when compound 1 was incubated with Aroclor-induced rat or human liver S9 supplemented with NADPH and CDP-choline, the metabolite profile was identical with that observed from the incubation with NADPH as the sole cofactor.
The in vitro metabolism of compound 2 was also investigated in incubations with liver S9 and primary hepatocytes. One minor hydroxylation product, 2-1, was observed across the in vitro samples.
In Vitro Formation of an Everolimus Conjugate.
When everolimus was incubated with Aroclor-induced rat or human liver S9, a conjugate (P + 165 Da) was detected by LC/UV/MS. The same product was also observed in incubations with the rat hepatocytes. The structural assignment of this product is described below in the MS section.
In Vivo Metabolism.
The metabolite profile of compound 1 was then characterized after intravenous administration (2 mg/kg) to BDC rats. In the bile samples collected between 0 and 24 hours, compound 1 represented only approximately 1% of the total drug-related material recovered from these matrices. The most abundant metabolites of compound 1 in bile were 1-1 (27%), the hydroxylation product also detected in the liver S9 and hepatocyte incubations, and 1-3 (52%), a metabolite that exhibited a net molecular weight addition of 181 Da relative to compound 1. Multiple minor metabolites were also formed along with pathways involving either mono-hydroxylation (1%, 1-2), bis-hydroxylation (1%, one product), or hydroxylation (mono- or bis-) accompanied by dehydrogenation (13% total, four products). In plasma, compound 1 represented the most abundant drug-related material (>76%) with presence of 1-1 and 1-2.
Separately, only 5% of the total drug-related material was attributed to unchanged compound 2 in bile samples (0‒24 hours) collected from BDC rats receiving an intravenous administration of 2 mg/kg. Compound 2 was converted to 2-1 as a major hydroxylation product (55%), along with multiple minor metabolites that were formed by either mono-hydroxylation (2-2, 6%), bis-hydroxylation (4%, one product), or hydroxylation and dehydrogenation (18% total, four products). Another product, 2-3, exhibiting a mass shift of 181 Da, represented a minor metabolite (6%) in bile. The conjugates (P + 181 Da) of compounds 1 and 2 were only detected in BDC rat bile and not in plasma, urine, or feces. No metabolite was detected in plasma.
Based on the total UV peak areas of the drug-related materials in bile, urine, and feces as well as the volumes of these three matrices, the disposition of compounds 1 and 2 was accomplished mainly via metabolism followed by excretion in bile (data not shown). Mass spectrometric analysis and structural assignments of these products are described in the next section.
MS Analysis.
The MSe product-ion spectrum of the protonated compound 1 (m/z 570) showed that fragmentation occurred on the bicyclo[1.1.1]pentan-1-ylcarbamoyl moiety, resulting in neutral losses of the bicyclic group (66 Da) and the amino-bicyclic group (83 Da) to yield the two most abundant product ions of m/z 504 and 487, respectively (Fig. 2A). The ion of m/z 504 further shed a CH2 group (14 Da) to give rise to the product ion of m/z 473. The product ion of m/z 427 was derived from a loss of the bicyclo[1.1.1]pentan-1-ylcarbamoyl moiety along with loss of CH2 (14 Da) and HF (20 Da) groups.
High-energy LC/MSe product ion spectra of compound 1 (A) and its two metabolites, 1-1 (B) and 1-3 (P + 181 Da) (C), acquired using a Waters Xevo QTOF mass spectrometer. The proposed fragment ions and corresponding theoretical masses are shown.
During MS/MS, the protonated 1-1 (m/z 586) experienced a neutral loss of one molecule of H2O (18 Da) to form a product ion of m/z 568 in the MSe spectrum (Fig. 2B), confirming it as a hydroxylation product. The diagnostic base peak of m/z 487 and an ion of m/z 504 together indicated hydroxylation of the bicyclic moiety. Two minor product ions of m/z 459 and 430 were proposed to result from a neutral loss of the hydroxyl-bicyclo[1.1.1]pentan-1-ylcarbamoyl moiety and subsequent loss of NHCH2 (29 Da), respectively. A second hydroxylation product (1-2) exhibited a fragmentation pattern similar to 1-1 (data not shown), suggesting that it was also a bicyclohydroxylation product.
Similarly, collision-induced dissociation of compound 2 and its hydroxylation product (2-1) occurred on the bicyclo[1.1.1]pentan-1-ylcarbamoyl moiety, giving rise to m/z 469 as the most abundant product ion in the MSe spectra of both compounds (Fig. 3, A and B). Thus, 2-1 was interpreted to have undergone hydroxylation on the bicyclic structure. In addition, 2-2, a hydroxylation product, showed a similar product ion spectrum to 2-1 and was proposed to be converted by bicyclohydroxylation (data not shown). Compounds 2 and 2-1 also gave rise to similar fragmentation patterns as 1 and its hydroxylation products (Fig. 2, A and B).
High-energy LC/MSe product ion spectra of compound 2 (A) and its two metabolites, 2-1 (B) and 2-3 (P + 181 Da) (C), acquired using a Waters Xevo QTOF mass spectrometer. The proposed fragment ions and their theoretical masses are shown.
Compound 1-3 (m/z 751, P + 181 Da), the unexpected metabolite of compound 1 in vivo, exhibited a base peak of m/z 184 in its MS/MS product ion spectrum (Fig. 2C). The presence of a fragment ion of m/z 487, which was also observed in the MS/MS spectrum of compound 1, suggested that the entire metabolic addition was associated with the aminobicyclic group. Likewise, the MS/MS analysis of 2-3 revealed fragment ions of m/z 184 and m/z 469, with the latter product ion indicating metabolic modification of the bicyclic portion (Fig. 3C). To elucidate the structural features of the added group, MSn experiments were carried out using an LTQ-Orbitrap mass spectrometer, and MS3 and MS4 product ion spectra of the ion of m/z 184 were collected (Fig. 4). The exact mass of the added moiety (184.0736 Da) and observed mass defect suggested an elemental composition of C5H15NO4P (Sleno, 2012). The MS3 ion spectrum exhibited two ions of m/z 60.0808 and 86.0965, corresponding to elemental compositions of C3H10N+ and C5H12N+, suggesting alkylamines characterized by different carbon chain lengths. The mass of another fragment ion of m/z 125.0000 was consistent with the elemental composition of an ethylphosphoric acid (C2H6O4P+). The MS4 spectrum of the m/z 125 ion (from the MS3 spectrum) gave rise to another ion of m/z 98.9842, consistent with the molecular weight of a phosphoric acid (H4O4P+).
The product ion spectra of the fragment ion of m/z 181.0022 (MS3; A) and the fragment ion of m/z 125.0000 (MS4; B) selected from the product ion spectrum (LC/MS/MS) of 1-3. The spectra were acquired using a Thermo LTQ-Orbitrap mass spectrometer. The proposed fragment ions and theoretical masses are shown.
Approximately 95% and 83% of biliary metabolites of compounds 1 and 2, respectively, were formed via metabolism of the bicyclopental moiety, as evidenced by the observation of the diagnostic fragment ions of m/z 487 or m/z 469 in their MS/MS product ion spectra (data not shown). On the other hand, the remaining in vivo metabolites of compounds 1 and 2 were proposed to form by metabolism on the remaining portions of the molecules based on their product ion spectra (data not shown) and are beyond the scope of this article. However, the relative abundances of these products are summarized in Table 1.
The MS/MS product ion spectrum of the everolimus conjugate (m/z 1123, P + 165 Da) exhibited product ions resulting from the fragmentation of both everolimus and the conjugate moiety (Supplemental Fig. 1). The ions of m/z 1091 and 1073 were formed by a neutral loss of methanol and further loss of water, respectively. The ions of m/z 829 and 532 were formed by fragmentation of the everolimus moiety as illustrated in Supplemental Fig. 1. In addition, multiple products ions could be ascribed to fragmentation of the conjugate moiety. The ions of m/z 184 and 86 were identical to those observed in the product ion spectra of 1-3 and 2-2 mentioned previously. The product ion spectrum was consistent with that of the everolimus POPC conjugate detected in human and multiple preclinical species (Zollinger et al., 2008).
NMR Spectroscopy.
The 1H and 13C chemical shifts of compound 1 and its respective conjugate (1-3), isolated from the BDC rat bile, were obtained during the acquisition of 1H, HMQC, and HMBC spectra and were assigned as summarized in Table 2. The eight protons of the aromatic rings, including those of the two fluorophenyl groups (H11/15, H12/14, H23, H27, and H26) and the furo[2,3-b]pyridine (H9), in compound 1 and 1-3 were characterized by no, or minor, changes in chemical shifts. No extra aromatic proton(s) was observed. In addition, the aliphatic protons of the N-methyl group (H30) and trifluoropropyl group (H18 and H19) also remained intact. On the bicyclic moiety, 1H chemical shifts of the six identical methylene protons (H38/40/41) and the single methine proton (H39) of compound 1 were 2.20 and 2.49 ppm, respectively. The peak area ratio between these signals was approximately 6:1 (data not shown). However, only one group of the bicyclic protons in 1-3 (2.47 ppm) were observed, and their NMR peak area was approximately 3-fold greater than that of the H18 (3.05 ppm) or H19 (2.66 ppm) methylene protons (data not shown). Furthermore, a long-range C-H correlation study (HMBC) of 1-3 revealed three distinct correlations between the bicyclic protons (2.47 ppm) and the bicyclic carbons of 42.9 ppm, 56.7 ppm, and 63.5 ppm (Fig. 5; Table 2). In addition, the bicyclic protons (2.47 ppm) exhibited a correlation with the bicyclic carbons (56.7 ppm) in the HMQC spectrum (Fig. 6), indicating direct connectivity.
Proton (1H NMR) and carbon chemical shifts (13C NMR for parent, and HMQC and HMBC for conjugates) of Compound 1 and 1-3 (P + 181 Da)
The proposed structure of the conjugate is shown below. The numbering system is for illustrative purposes only and does not correspond to International Union of Pure and Applied Chemistry nomenclature.
HMBC spectrum of 1-3, illustrating long-range correlation of 1H and 13C nuclei.
HMQC spectrum of 1-3, showing two-dimensional heteronuclear chemical shift correlations between directly bonded 1H and 13C nuclei.
The 1H spectrum of 1-3 also showed three additional and distinct groups of aliphatic signals at 3.24, 3.65, and 4.28 ppm (Table 2). The peak area ratio among these signals was approximately 9:2:2 (data not shown). The HMQC spectrum demonstrated direct connectivity between these protons (3.24, 3.65, and 4.28 ppm) and carbons at 54.5, 67.3, and 60.3 ppm, respectively (Fig. 6), and COSY was used to show a correlation between the protons of 3.65 ppm and 4.28 ppm (Fig. 7). Moreover, the HMBC experiment established three long-range C-H correlations: the 1H of 3.24 ppm and 13C of 54.5 ppm, the 1H of 3.24 ppm and 13C of 67.3 ppm, and the 1H of 3.65 ppm and 13C of 54.5 ppm (Fig. 5).
COSY spectrum of 1-3, depicting through-bond 1H-1H correlations.
Discussion
The metabolite profiles of compounds 1 and 2 in the BDC rat studies demonstrated extensive metabolism of both molecules in the liver and pointed to the bicyclopentane moiety as the main site of metabolism. In addition to hydroxylation, with or without dehydrogenation, metabolism of the bicyclic group also yielded unexpected products characterized by a mass addition of 181 Da, which did not match any of the metabolic mass shifts monitored routinely in our laboratory.
The MS/MS product ion spectra of the unknown products, 1-3 and 2-3, included a base peak of m/z 184, whose multistage MSn fragmentation spectra assisted in the characterization of multiple structural features of the added moiety (Fig. 4). The m/z 60.0808 and m/z 86.0965 ions were in agreement with elemental compositions of alkyl amines, C3H10N and C5H12N, respectively, whereas the ions of m/z 98.9842 and m/z 125.0000 matched the masses of phosphoric acid (H4O4P+) and ethylphosphoric acid (C2H6O4P+), respectively. Therefore, the added moiety appeared to consist of an ethyl-substituted phosphoric acid connected to an alkyl amine, with the degree of substitution at the nitrogen atom still undetermined. All elemental compositions listed above represented the theoretical formulas that were closest to the observed masses, providing additional support for the structure assignments.
To elucidate the structure of the added moiety, 1-3 was isolated from rat bile and analyzed by NMR. The 1H spectra indicated that the metabolic change did not occur on any of the aromatic carbons, N-methyl group or trifluoropropyl group, consistent with the MS/MS analysis implicating the bicyclic group as the likely soft spot. Comparison of the 1H spectra describing the bicyclic portions of compound 1 and 1-3 led to one of the most important findings: only one group of six equivalent protons (2.47 ppm) in 1-3 was observed, in contrast with the two groups of protons (2.20 and 2.49 ppm) that were apparent in compound 1. This observation excluded the possibility of modification of one of the three equivalent methylene carbons (C38/C40/C41) which would have otherwise resulted in three different groups of protons—one H39 proton, four equivalent methylene protons, and the one proton corresponding to the site of metabolism. Therefore, metabolism was shown to occur at the bridgehead carbon (C39), causing the six equivalent protons H38/H40/H41 to shift downfield (2.47 ppm) and leaving no proton connected to C39. Moreover, the chemical shift of C39 of 1-3, which was assigned as 63.5 ppm, exhibited a large downfield shift from 26.1 ppm, providing clear evidence that C39 was connected to the added moiety through an oxygen.
The conjugated group also left a diagnostic footprint in the aliphatic region of the NMR spectra. First, the characteristic NMR peak area ratio of the three distinct protons, 9 (3.24 ppm), 2 (3.65 ppm), and 2 (4.28 ppm), suggested the presence of three equivalent methyls and two different methlyene groups. Second, the HMQC and COSY spectra supported the notion that the protons of 3.65 ppm and 4.28 ppm were attached to neighboring carbons (67.3 and 60.3 ppm, respectively)—likely an ethylene, as suggested by the MSn results. Furthermore, the HMQC and HMBC spectra provided evidence of a direct connection and through-bond proximity between the carbon of 54.5 ppm and the nine equivalent protons of 3.24 ppm, suggesting the presence of three methyl groups likely attached to a nitrogen. Collectively, the LC/MS/MS and NMR data supported a structure assignment of 1-3 in which a POPC group [H2PO4-CH2CH2N(CH3)3] was connected to the methine carbon (C39) of the bicyclic moiety through an ester bond.
It is well known that CPT catalyzes the final step of the Kennedy reaction by adding POPC from CDP-choline to a diacylglycerol (DAG), yielding phosphatidylcholine, a major component of biologic membranes including microsomal membranes of all eukaryotes (Blank et al., 1979; Demopoulos et al., 1979; McMaster and Bell, 1997). The membrane-bound CPT resides in the endoplasmic reticulum, a primary location for the biosynthesis of nearly all kinds of phospholipids from fatty acyl CoA and glycerol 3-phosphoate (Bishop and Bell, 1988; Daleke, 2003; Vance and Vance, 2008; Lagace and Ridgway, 2013). The substrate, DAG, docks in the active site of CPT with its hydroxyl group exposed to the cytosol, in which the cofactor, CDP-choline, is present (McMaster and Bell, 1997; Alberts et al., 2008). In our study, the POPC conjugation of compounds 1 and 2 was presumably catalyzed by CPT after hydroxylation of the bicyclic methine, a prerequisite reaction that was catalyzed by P450. It is reasonable that, after being released by P450, the tertiary alcohol product of compound 1 or 2 binds to the CPT active site in the endoplasmic reticulum membrane, with the newly added hydroxyl group facing the cytosol, and is then conjugated by addition of a POPC to the hydroxyl group.
Although mono- and bis-hydroxylation of the bicyclic moiety of compound 1 or 2 led to the formation of multiple metabolites, only one mono-hydroxylation product appeared to be a suitable substrate of CPT, indicating a degree of selectivity on the part of the enzyme. In addition, replacement of the fluorophenyl ring with an unsubstituted phenyl as in compound 2, a subtle structural change relative to compound 1, resulted in a markedly diminished extent of POPC conjugation, also reflecting the substrate specificity of the enzyme. On the other hand, the hydroxylation products of compound 1 or 2 do not appear to be closely related to the structures of the endogenous substrates of CPT, which include DAGs possessing varied lengths of saturated or unsaturated fatty acid chains (usually 16–22 carbons) and a primary alcohol group, demonstrating that CPT can recognize molecules whose chemical scaffold differs from those of its endogenous substrates. Previously, Zollinger et al. (2008) reported that everolimus, a potent immunosuppressant, formed a POPC ester via direct conjugation of a hydroxyl group as a prominent metabolite in rats, mice, cynomolgus monkeys, and humans. Formation of POPC conjugates of everolimus was also unexpected in light of the structural differences between DAGS and everolimus.
In our study, the formation of the everolimus POPC conjugate was confirmed in incubations with liver S9 fortified with CDP-choline and with rat hepatocytes, demonstrating the activity of CPT in these in vitro systems. However, no POPC product was observed after an attempt to recapitulate formation of the POPC conjugate of compound 1 in Aroclor-induced rat liver S9 supplemented with NADPH and CDP-choline, nor were such conjugates formed in incubations with hepatocytes despite the presence of all enzymes and cofactors required to carry out this two-step reaction. The apparent discrepancy between the in vitro and in vivo results is explained by inadequate turnover to the appropriate tertiary alcohol product during the first of these two steps in vitro (Table 1). On the other hand, this was not an issue in the everolimus example, since it features a primary alcohol and acts as a direct CPT substrate without the need for phase I biotransformation. Whereas the synthesis of the methine-hydroxylation products of compound 1 or 2 and the subsequent incubation of these metabolites in vitro might have facilitated the detection of the POPC conjugates and allowed this disconnect to be resolved fully, the deprioritization of the bicyclic series in the medicinal chemistry program precluded the availability of the requisite resources to address this question.
Whether the hydroxylation products of compounds 1 and 2 would be converted to the POPC conjugates in human subjects remains unknown. The observation of the POPC conjugates only in bile from the BDC rats, as well as the structural dissimilarity between these antiviral compounds and the endogenous substrates of CPT, make it difficult to predict whether and to what extent the analogs of compounds 1 and 2 would undergo POPC conjugation in vivo without conducting additional experiments.
From the standpoint of program decision making, the observation of this unusual reaction in BDC rats was accompanied by uncertainty over the possibility that the POPC-xenobiotic conjugate formation might interfere with the de novo synthesis and function of phospholipids. However, the results of these studies did not allow the magnitude of this effect to be estimated, since the BDC rat study was not designed to measure mass balance or determine the flux through any specific metabolic pathway in vivo. Although substitution of the methine of the bicyclic group was considered as a means of blocking POPC conjugation, this idea was not pursued given the difficulty in predicting whether hydroxylation and POPC conjugation might still occur on another site within the bicyclic group.
In summary, we observed an uncommon POPC conjugation pathway associated with HCV NS5B inhibitors featuring a bicyclic moiety. The LC/MSn and NMR analysis was used to identify the bicyclopental methine carbon as the site that underwent sequential hydroxylation and esterification with POPC. The observation of this reaction in vivo, but not in vitro, complicated our routine efforts to establish relationships across species and anticipate what might happen in a clinical setting. These findings stand as another example of how enzymes with seemingly exclusive functions in the transformation of endogenous substrates may, in certain circumstances, get recruited to participate in the metabolism of xenobiotics.
Acknowledgments
The authors thank Nicholas Meanwell for manuscript review and valued discussion, and Dawn D. Parker, Jennifer G. Pizzano, and Jean Simmermacher-Mayer for technical assistance.
Authorship Contributions
Participated in research design: Zhuo, Cantone, Wang, Drexler, Mosure.
Conducted experiments: Zhuo, Cantone, Wang, Leet.
Contributed new agents or analytic tools: Zhuo, Cantone, Wang, Drexler, Yueng, Eastman, Parcella, Kadow.
Performed data analysis: Zhuo, Cantone, Wang, Leet, Drexler, Huang, Johnson.
Wrote or contributed to the writing of the manuscript: Zhuo, Wang, Leet,
Drexler, Soars, Johnson.
Footnotes
- Received December 18, 2015.
- Accepted March 7, 2016.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- BDC
- bile duct–cannulated
- BMS-791325
- (1aR,12bS)-8-cyclohexyl-N-(dimethylsulfamoyl)-11-methoxy-1a-{[(1R,5S)-3-methyl-3,8-diazabicyclo[3.2.1]oct-8-yl]carbonyl}-1,1a,2,12b-tetrahydrocyclopropa[d]indolo[2,1-a][2]benzazepine-5-carboxamide
- CDP-choline
- cytidine-diphosphocholine
- compound 1
- 5-(3-(bicyclo[1.1.1]pentan-1-ylcarbamoyl)-4-fluorophenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide
- compound 2
- 5-(3-(bicyclo[1.1.1]pentan-1-ylcarbamoyl)phenyl)-2-(4-fluorophenyl)-N-methyl-6-(3,3,3-trifluoropropyl)furo[2,3-b]pyridine-3-carboxamide COSY, 1H-1H homonuclear correlation spectroscopy
- CPT
- CDP-choline:1,2-diacylglycerol cholinephosphotransferase
- DAG
- diacylglycerol
- DMSO
- dimethyl sulfoxide
- HCV
- hepatitis C virus
- HMBC
- heteronuclear multiple bond correlation
- HMQC
- heteronuclear multiple quantum coherence
- HPLC
- high-pressure liquid chromatography
- LC
- liquid chromatography
- MS
- mass spectrometry
- MS/MS
- tandem mass spectrometry
- MSn
- multiple-stage mass spectrometry
- m/z
- mass-to-charge ratio
- NADPH
- β-nicotinamide adenine dinucleotide phosphate tetrasodium salt
- NMR
- nuclear magnetic resonance
- NS
- nonstructural protein
- P450
- cytochrome P450
- POPC
- phosphocholine
- Copyright © 2016 by The American Society for Pharmacology and Experimental Therapeutics