Biotransformation and Pharmacokinetic Evaluation of PF-5190457, A Novel Drug Candidate for Alcoholism

Alcohol use disorder is a chronic condition characterized by an inability to control alcohol intake. Recent research has shown that appetite-regulating hormones, ghrelin, and leptin, may play a role in alcohol craving. In addition, it is known that food and alcohol craving have overlapping receptors in the brain reward system. Ghrelin, a naturally occurring hormone in the human body, is produced mainly by ghrelinergic cells of the gastric mucosa. Once the hormone crosses the blood-brain barrier, it binds to growth hormone secretagogue receptor (ghrelin receptor) in the hypothalamus and stimulates the appetite and food intake. Leptin, a hormone produced from adipose tissue has an opposite effect to that of ghrelin and reduces hunger. Leptin and ghrelin work in tandem to maintain energy homeostasis in the body. It was demonstrated that antagonizing ghrelin receptor reduces alcohol craving thus making it a potential pharmacological target to treat alcohol use disorders. PF-5190457 is a highly selective, orally bioavailable ghrelin receptor inverse agonist developed by Pfizer. PF-5190457 was made available for research by the National Center for Advancing Translational Sciences (NCATS) under the NIH New Therapeutic Uses Pilot Program. Initial studies by Pfizer had shown that the compound is safe and well tolerated in healthy volunteers after oral administration. To date, the compound has completed Phase 1b clinical trials in nontreatment-seeking patients with alcohol use disorder. The plasma samples obtained from the study was evaluated for the pharmacokinetics of PF-5190457 and its major hydroxy metabolite. In order to measure the concentration of the drug and its metabolite, a liquid chromatography tandem mass spectroscopy method (LC-MS/MS) was developed and validated as described in Manuscript I (to be submitted to Journal of Chromatography B: Biomedical Sciences and Applications). The manuscript describes a new LC-MS/MS method developed for this purpose and the assay was fully validated according to the FDA guidance document for bioanalytical methods. The analytes of interest were extracted from plasma with methanol using a simple protein precipitation technique, and tacrine was used as the internal standard. All the bioanalysis method parameters, i.e., sensitivity, specificity, linearity, accuracy, precision, matrix effect, recovery, and stability were within the recommended FDA guidelines. Incurred sample reanalysis (10% of the patient plasma samples) was performed to evaluate the repeatability of the new method and was found to be within the acceptance criteria. Manuscript II (to be submitted to Drug Metabolism and Disposition) describes the biotransformation of PF-5190457 (molecular weight: 513). Before this work in our lab, no information on the metabolism of the PF-5190457 in humans was available. Extensive LC-MS/MS profiling of plasma from Phase 1b study showed many circulating metabolites. Subsequent analysis revealed the presence of a significant hydroxy metabolite (m/z 529) along with minor glucuronide and hydroxy glucuronide metabolites. Further exploration of the fragmentation pattern of the major hydroxy metabolite (m/z 351 and m/z 225) suggested possible hydroxylation between the two nitrogen atoms of the drug. In parallel, in vitro experiments were conducted in human hepatic microsomes and cytosolic fractions to characterize the drug metabolism enzymes responsible for the formation of metabolites. Preliminary data revealed the formation of hydroxy metabolite in the cytosol but not in the microsomes. The reaction was found to be independent of the presence of nicotinamide adenine dinucleotide phosphate (NADPH) in the incubation media. Addition of a non-selective CYP450 inhibitor (1-amino benzotriazole) inhibited the formation of all the minor metabolites but not the hydroxy metabolite. A comparative mass spectra analysis showed the presence of hydroxy metabolite in the human liver cytosol incubations which was previously found in human plasma. PF-5190457 was also incubated in hepatocytes leading to the formation of major and minor metabolites. Further addition of an aldehyde oxidase inhibitor (hydralazine) in hepatocytes completely inhibited the formation of the hydroxy metabolite indicating the involvement of aldehyde oxidase in the metabolism of the PF-5190457. The hydroxy metabolite was also found in incubations conducted with PF-5190457 and cytosol from different animal species. The aldehyde oxidase metabolite of PF-5190457 was biosynthesized, from female mouse liver cytosol and the structure of the metabolite was confirmed using NMR analysis indicating the hydroxylation position to be on the pyrimidine ring between the two nitrogen atoms. Literature shows both aldehyde and xanthine oxidase share substrate specificity. Specific inhibitors of both the enzymes were used to identify their involvement in the metabolism of the drug. Raloxifene and febuxostat were used as selective aldehyde oxidase and xanthine oxidase inhibitor, in human liver cytosol incubations with PF-5190457 respectively. The data from inhibition study suggested a role for both the enzymes in the biotransformation of the drug with aldehyde oxidase enzyme being dominant over xanthine oxidase. The kinetic parameters Km and Vmax for hydroxy metabolite formation in human liver cytosol were 42 ± 4 μM and 0.12 ± 0.003 nmol/min/mg protein, respectively. Aldehyde oxidase (AOX1) is gaining increased attention for its role in the metabolism of new drug candidates. A significant variability in the levels of the AOX1 protein expression in the human liver has been reported in published studies. We assessed the AOX1 levels in a novel repository of human liver (n=104) using mass spectroscopy as described in Manuscript III (to be submitted to Drug Metabolism and Disposition). The levels of AOX1 were estimated using traditional data-dependent acquisition and upcoming SWATHMS method. A strong and significant correlation (Spearman r-0.72, P<0.0001) was observed between the two methods making a case for the use of SWATH based methods in the field of drug metabolism. Expression levels of AOX1 determined in our liver bank using MaxQuant (proteomics data analysis software) were found to be well within the literature reported range (2 49 pmol/mg total cytosolic protein). Further, we evaluated the effect of demographic factors (gender, age, ethnicity, drug usage, alcohol consumption, and smoking) and disease conditions (diabetes, non-alcoholic fatty liver disease) on the AOX1 expression. We found no significant association between demographic factors with the expression of the AOX1 protein. Similarly, disease conditions were not found to influence the expression of AOX1 indicating other unidentified factors influencing the enzyme expression. Manuscript IV (to be submitted to British Journal of Clinical Pharmacology) outlines the pharmacokinetics data via non-compartmental analysis of PF-5190457 and its hydroxy metabolite. The concentrations of analytes from phase 1b trials were evaluated in N=12 patients at placebo, 50 and 100 mg doses using validated LC-MS/MS method as described in Manuscript I. The half-life of the metabolite (~13 h) was found to be longer than the parent drug (~6 h). Additionally, the rate of formation of hydroxy metabolite was studied in cytosol prepared from 10 human liver samples. The rate of formation of hydroxy metabolite was found to be highly variable with a 20.5-fold (0.02 – 0.41 nmol/min/mg protein) variation among the ten samples. Interestingly, the AOX1 protein expression for same samples showed only 2.5-fold variability (16.8 44.1 pmol/mg protein). In summary, an understanding of the enzymes responsible for the metabolism of a drug is essential to avoid drug-drug interactions or the presence of polymorphic enzymes interfering with the therapeutic outcome of the drug. Collectively, we for the first time determined the clinical pharmacokinetics of PF5190457 and its major metabolite (hydroxy metabolite) in alcoholic patients. Further studies suggested the involvement of aldehyde oxidase (major role) and xanthine oxidase (minor role) in the metabolism of the PF-5190457. Moreover, to examine the variability in the protein expression of the enzymes, we quantified AOX1 in a large liver databank. This work also makes a case for considering non-cytochrome-P450 enzymes in the early screening during drug discovery. Considering the longer half-life of metabolite than the parent, it will be warranted to investigate the pharmacological properties of hydroxy metabolite in future studies.

crosses the blood-brain barrier, it binds to growth hormone secretagogue receptor (ghrelin receptor) in the hypothalamus and stimulates the appetite and food intake. Leptin, a hormone produced from adipose tissue has an opposite effect to that of ghrelin and reduces hunger. Leptin and ghrelin work in tandem to maintain energy homeostasis in the body. It was demonstrated that antagonizing ghrelin receptor reduces alcohol craving thus making it a potential pharmacological target to treat alcohol use disorders.
PF-5190457 is a highly selective, orally bioavailable ghrelin receptor inverse agonist developed by Pfizer. PF-5190457 was made available for research by the National Center for Advancing Translational Sciences (NCATS) under the NIH New Therapeutic Uses Pilot Program. Initial studies by Pfizer had shown that the compound is safe and well tolerated in healthy volunteers after oral administration. To date, the compound has completed Phase 1b clinical trials in nontreatment-seeking patients with alcohol use disorder. The plasma samples obtained from the study was evaluated for the pharmacokinetics of PF-5190457 and its major hydroxy metabolite. In order to measure the concentration of the drug and its metabolite, a liquid chromatography tandem mass spectroscopy method (LC-MS/MS) was developed and validated as described in Manuscript I (to be submitted to Journal of Chromatography B: Biomedical Sciences and Applications). The manuscript describes a new LC-MS/MS method developed for this purpose and the assay was fully validated according to the FDA guidance document for bioanalytical methods. The analytes of interest were extracted from plasma with methanol using a simple protein precipitation technique, and tacrine was used as the internal standard. All the bioanalysis method parameters, i.e., sensitivity, specificity, linearity, accuracy, precision, matrix effect, recovery, and stability were within the recommended FDA guidelines. Incurred sample reanalysis (10% of the patient plasma samples) was performed to evaluate the repeatability of the new method and was found to be within the acceptance criteria.
Manuscript II (to be submitted to Drug Metabolism and Disposition) describes the biotransformation of PF-5190457 (molecular weight: 513). Before this work in our lab, no information on the metabolism of the PF-5190457 in humans was available. Extensive LC-MS/MS profiling of plasma from Phase 1b study showed many circulating metabolites. Subsequent analysis revealed the presence of a significant hydroxy metabolite (m/z 529) along with minor glucuronide and hydroxy glucuronide metabolites. Further exploration of the fragmentation pattern of the major hydroxy metabolite (m/z 351 and m/z 225) suggested possible hydroxylation between the two nitrogen atoms of the drug. In parallel, in vitro experiments were conducted in human hepatic microsomes and cytosolic fractions to characterize the drug metabolism enzymes responsible for the formation of metabolites. Preliminary data revealed the formation of hydroxy metabolite in the cytosol but not in the microsomes. The reaction was found to be independent of the presence of nicotinamide adenine dinucleotide phosphate (NADPH) in the incubation media. Addition of a non-selective CYP450 inhibitor (1-amino benzotriazole) inhibited the formation of all the minor metabolites but not the hydroxy metabolite. A comparative mass spectra analysis showed the presence of hydroxy metabolite in the human liver cytosol incubations which was previously found in human plasma. PF-5190457 was also incubated in hepatocytes leading to the formation of major and minor metabolites. Further addition of an aldehyde oxidase inhibitor (hydralazine) in hepatocytes completely inhibited the formation of the hydroxy metabolite indicating the involvement of aldehyde oxidase in the metabolism of the PF-5190457. The hydroxy metabolite was also found in incubations conducted with PF-5190457 and cytosol from different animal species. The aldehyde oxidase metabolite of PF-5190457 was biosynthesized, from female mouse liver cytosol and the structure of the metabolite was confirmed using NMR analysis indicating the hydroxylation position to be on the pyrimidine ring between the two nitrogen atoms. Literature shows both aldehyde and xanthine oxidase share substrate specificity. Specific inhibitors of both the enzymes were used to identify their involvement in the metabolism of the drug. Raloxifene and febuxostat were used as selective aldehyde oxidase and xanthine oxidase inhibitor, in human liver cytosol incubations with PF-5190457 respectively. The data from inhibition study suggested a role for both the enzymes in the biotransformation of the drug with aldehyde oxidase enzyme being dominant over xanthine oxidase. The kinetic parameters Km and Vmax for hydroxy metabolite formation in human liver cytosol were 42 ± 4 µM and 0.12 ± 0.003 nmol/min/mg protein, respectively.
Aldehyde oxidase (AOX1) is gaining increased attention for its role in the metabolism of new drug candidates. A significant variability in the levels of the AOX1 protein expression in the human liver has been reported in published studies. We assessed the AOX1 levels in a novel repository of human liver (n=104) using mass spectroscopy as described in Manuscript III (to be submitted to Drug Metabolism and Disposition). The levels of AOX1 were estimated using traditional data-dependent acquisition and upcoming SWATH-MS method. A strong and significant correlation P<0.0001) was observed between the two methods making a case for the use of SWATH based methods in the field of drug metabolism. Expression levels of AOX1 determined in our liver bank using MaxQuant (proteomics data analysis software) were found to be well within the literature reported range (2 -49 pmol/mg total cytosolic protein). Further, we evaluated the effect of demographic factors (gender, age, ethnicity, drug usage, alcohol consumption, and smoking) and disease conditions (diabetes, non-alcoholic fatty liver disease) on the AOX1 expression. We found no significant association between demographic factors with the expression of the AOX1 protein. Similarly, disease conditions were not found to influence the expression of AOX1 indicating other unidentified factors influencing the enzyme expression.

Manuscript IV (to be submitted to British Journal of Clinical
Pharmacology) outlines the pharmacokinetics data via non-compartmental analysis of PF-5190457 and its hydroxy metabolite. The concentrations of analytes from phase 1b trials were evaluated in N=12 patients at placebo, 50 and 100 mg doses using validated LC-MS/MS method as described in Manuscript I.
The half-life of the metabolite (~13 h) was found to be longer than the parent drug (~6 h). Additionally, the rate of formation of hydroxy metabolite was studied in cytosol prepared from 10 human liver samples. The rate of formation of hydroxy metabolite was found to be highly variable with a 20.5-fold (0.02 -0.41 nmol/min/mg protein) variation among the ten samples. Interestingly, the AOX1 protein expression for same samples showed only 2.5-fold variability (16.8 -44.1 pmol/mg protein).
In summary, an understanding of the enzymes responsible for the metabolism of a drug is essential to avoid drug-drug interactions or the presence of polymorphic enzymes interfering with the therapeutic outcome of the drug.
Collectively, we for the first time determined the clinical pharmacokinetics of PF-5190457 and its major metabolite (hydroxy metabolite) in alcoholic patients.
Further studies suggested the involvement of aldehyde oxidase (major role) and xanthine oxidase (minor role) in the metabolism of the PF-5190457. Moreover, to examine the variability in the protein expression of the enzymes, we quantified AOX1 in a large liver databank. This work also makes a case for considering non-cytochrome-P450 enzymes in the early screening during drug discovery.
Considering the longer half-life of metabolite than the parent, it will be warranted xiii        Abstract PF-5190457 is a selective and potent ghrelin receptor inverse agonist presently undergoing clinical trials to treat alcoholism. We have established a selective and sensitive liquid chromatography-tandem mass spectrometry method for determination of PF-5190457 and its hydroxy metabolite (PF-6870961) in human plasma and validated the assay according to FDA guidelines. Plasma samples were prepared by simple protein precipitation in methanol using tacrine as an internal standard (IS). Chromatographic separation was carried out on an Acquity UPLC BEH C18 (2.1mm×50mm) column with 1.7 μm particle size and 130 Å pore size column at a flow rate of 0.25 mL/min by a gradient method with a runtime of 11.30 min. Detection was performed using multiple reaction monitoring (MRM) at positive ion mode with two fragments for each analyte.

LIST OF TABLES
Tacrine was found to be a suitable IS. The calibration curves were linear in the concentration range of 1-1000 ng/mL and 2-250 ng/mL for PF-5190457 and hydroxy metabolite respectively with a correlation coefficient ≥0.996. The interassay precision and accuracy results obtained were within the FDA recommended guidelines. The analytes were found to be stable under varied stability conditions. The validated method was utilized to measure plasma concentration of PF-5190457 and its hydroxy metabolite in clinical studies.

Introduction
PF-5190457 ( Fig. 1) is a novel peripherally acting inverse agonist of ghrelin receptor discovered by Pfizer pharmaceuticals. It is a small molecule with a molecular weight (Mw) 512.67 belonging to a member of spiro-azetidinopiperidine series [1,2]. Ghrelin, an endogenous peptide released from the stomach, is a ligand for growth hormone secretagogue receptor in the hypothalamus and regulates appetite and food intake in the body [3][4][5].
Clinical studies of PF-5190457 in healthy volunteers showed the drug is well-tolerated and exhibit linear pharmacokinetics [13]. Following oral administration, PF-5190457 is rapidly absorbed with a Tmax of 0.3-3 hour and its elimination half-life (t1/2 ) was ~8 h [13]. The metabolism pathway of PF-5190457 suggests cytosolic oxidative reaction involving molybdenum co-factor containing enzymes, aldehyde oxidase, and xanthine oxidase, leading to the formation of a hydroxy metabolite (Figure 1b) (Manuscript II). PF-5190457 is currently undergoing clinical trials for the treatment for alcohol use disorder and type 2 diabetes [8,14].
Previously developed validated UPLC-MS/MS method in the laboratory did not include the hydroxy metabolite in the assay [14]. The current manuscript demonstrates a simple and selective ultra performance liquid chromatographytandem mass spectroscopy (UPLC-MS/MS) for simultaneous determination of PF-5190457 and its hydroxy metabolite (PF-6870961) in human plasma using tacrine as an IS. The present method is established with the aim of determining PF-5190457 and its hydroxy metabolite in plasma samples obtained in clinical trials of patients with alcohol use disorder.

LC-MS/MS instrumentation, conditions and data processing
The liquid chromatography separation was performed using an Acquity UPLC system comprising with a binary pump and a built-in column heater. A 10 μL sample loop was used to inject samples in partial loop injection mode.
Chromatography separation of analytes and IS was accomplished within 11.30 min using gradient elution method using an Acquity UPLC BEH C18 ( was maintained for 2 min, B was increased from 2% to 80% over 8.30 min, and held till 9.30 min and returned to initial conditions, followed by a 2 min equilibration for a total run time of 11.30 min. Mass spectral analysis and quantification was carried out using multiple reaction monitoring (MRM) on a positive electrospray ionization mode with the following parameters: desolvation temperature of 350°C, desolvation gas flow 650 L/h, capillary voltage 3.50 kV and cone voltage 20 V. The fragmentation pattern for PF-5190457 and IS were similar to previously reported methods [14][15][16]. The optimized settings of mass spectrometer voltage and the retention time for analytes and internal standards with quantifier and qualifier fragments are summarized in Table 1.
Data acquisition and processing were performed using MassLynx™ software (V 4.1) and TargetLynx™ tool respectively. Calibration curves were constructed by calculating the analyte to IS peak area ratio against analyte concentrations.

Method validation
The validation of this assay was performed according to the requirements as outlined in guidelines established by the US Food and Drug Administration (US FDA) to evaluate the method with respect to selectivity, sensitivity, linearity, accuracy, precision, recovery, matrix effect, recovery and stability of analytes [17].

Specificity and Sensitivity
The method specificity was evaluated prior to the main validation batches for the presence of interference from endogenous compounds by screening blank plasma of six donors. The sensitivity of the assay was evaluated by analyzing blank plasma samples from six different donors and spiked plasma samples at the lower limit of quantification (LLOQ) levels of the analytes.

Linearity, accuracy, and precision
The linearity of the method was determined by analysis of standard plots associated with an eight-point calibration curve. Inter-assay precision and accuracy were calculated after the replicates in six different analytical runs. Peak area ratios of analytes to IS obtained from MRM were utilized for the construction of calibration curves, using least squares weighted (1/x) linear regression. Carry over test was assessed to demonstrate any carryover effect of analytes and IS.

Matrix effect and Recovery
The components present in the matrix (plasma) can interfere with ionization efficiency by coeluting with analytes of interest [18]. Matrix effect for ion suppression and enhancement was assessed using blank and spiked human plasma using post column infusion experiment as described previously [14,19].
The chromatograms can be visually evaluated to verify there are no interfering peaks at the elution time of analytes and IS. Likewise, the interference of phospholipids was determined as they are highly abundant endogenous compounds in biological membranes and can interfere with elution of analytes [20,21]. The major transitions of phospholipids were monitored with the precursor ion (Q1) at m/z 496, 522, 524, 758 and 782 and product ion (Q3) with m/z 184 in the method [22]. Recovery was determined at two concentration levels (LQC and HQC) by comparing the peak area of the analytes from the preextracted plasma standard with that obtained from post-extracted standard at the same concentration for the QC samples.

Stability
Stability of analytes were evaluated at two concentrations (LQC and HQC). The freeze-thaw stability was evaluated after undergoing three freeze thaw cycles, benchtop stability for 6 h and short-term stability was assessed after storage of the test samples at −80°C for 1 week. The autosampler storage stability was determined by storing the QC samples for ~48 h under autosampler condition (maintained at 10 C) before being analyzed.

Sample Reanalysis
The results from previously validated UPLC-MS/MS method for determination of PF-5190457 in patient plasma samples (n=45) by the laboratory was correlated with the current method to test the assay's ability to reproduce the data [14].

LC-MS assay
The goal of this work was to develop and validate a simple and sensitive assay method for the simultaneous determination of PF-5190457 and its hydroxy metabolite in human plasma. The MS optimization was performed by direct infusion of solutions of PF-5190547, hydroxy metabolite and IS into the ESI source of the mass spectrometer. The best signal for both the analytes and IS was achieved using a positive ion mode. Protonated adducts for PF-5190547 (m/z 513) and hydroxy metabolite (m/z 529) and their respective high intensity fragments (m/z 209 and m/z 225) were selected respectively for quantification.
Several trials for chromatographic conditions were performed especially to select an optimal mobile phase and analytical C18 columns. These trials determined that the present method achieved an optimal degree of chromatographical separation for simultaneous separation of different hydroxy metabolites as well as the parent drug.

Sample preparation optimization:
A simple protein precipitation technique was utilized for the extraction of analytes and IS from the plasma samples. Acetonitrile, methanol, DMSO, 50% DMSO, 50% acetonitrile, 50% methanol mixture were investigated as organic solvents for stock solution preparation. It was noticed that the metabolite was unstable in acetonitrile. Methanol was then chosen as the organic solvent for preparation of stock solution for the analytes and IS as well as only methanol was used in mobile phase.

Selection of internal standard
The stable isotopically labeled IS was not available for hydroxy metabolite and therefore we have demonstrated the suitability of readily available, costeffective compounds with no evidence of cross-talk with the analytes of interest in the current assay. Various compounds with similar LogP and Pka to that of analytes like indoprofen, chloroxazane, naproxen, tacrine and rifampicin were tested for selection of suitable IS for the assay. Tacrine was deemed to be the most suitable IS for quantification of PF-5190457 and hydroxy metabolite as it ionizes in the same polarity along with the respective analytes.

Specificity and selectivity
The determination of PF-5190547, hydroxy metabolite and IS at their respective retention time were not affected by interference from co-extracted matrix components under the present LC-MS/MS conditions used in blank plasma obtained from six different donors. Sensitivity was determined by comparing the baseline from blank plasma obtained from six different donors in comparison to that spiked with analytes at LLOQ. There was no interference observed from endogenous compounds. The representative extracted ion chromatograms of the analytes and the IS in human plasma are shown in Figure   2.

Linearity, Precision and Accuracy
Calibration was found to be linear over the concentration range of 1-1000 ng/mL for PF-5190547 and 2-250 ng/mL for hydroxy metabolite. The curve was fitted with linear regression using 1/x weighing function. The percentage deviation for PF-5190547 and hydroxy metabolite was less than 8.2% and 12.2% respectively ( for the hydroxy metabolite. There was no between-sample carryover detected following the double blank injections.

Matrix effect and Recovery
In the present study, the determination of PF-5190547 and hydroxy metabolite is not affected by co-extracted matrix components under the LC-MS/MS conditions used Figure 3a. There was no interference from the phospholipids on the elution of analytes Figure 3b. The percent mean recovery for PF-5190547 and hydroxy metabolite ranged from 95-103%.

Stability
The data representing the stability in plasma at two concentration levels (LQC and HQC) for PF-5190547 and hydroxy metabolite for bench top stability, three freeze and thaw cycles, autosampler and short-term stability are given in Table 4. These results indicate that the analytes are stable in human plasma as the mean of the results of the tested samples were within the acceptance criteria of ± 15% of the nominal concentrations. These findings suggest no stabilityrelated problems would be expected during pharmacokinetic studies and analysis.

Sample Reanalysis
The patient samples (n=45) previously analyzed by UPLC-MS/MS method and subsequently analyzed by the current method were within ± 15.2% for PF-5190457. The data generated by this method is reproducible despite differing the LC-MS/MS conditions. Method reproducibility was also determined by ISA (n=45) following the FDA Guidance for Industry for parent and metabolite. More than two-thirds of the sample were within ± 20%. The incurred sample analysis indicates that both parent and its metabolite are stable within human plasma from subjects. Bland-Altman plot is recommended over Pearson's or Spearman's correlation coefficient for correlation analysis for ISA as the range of data can influence the correlation [23]. The correlation between both the LC-MS/MS methods and ISA for PF-5190457 and hydroxy metabolite are depicted via Bland-Altman plots in Figure 4(a)-(c).

Method Application
The present validated method has been successfully applied to measure the concentration of PF-5190547 and its hydroxy metabolite after oral  Figure 5 represents plasma concentration time profile after an oral dose of 50 mg dose in a study subject.

Conclusion
In this manuscript, we describe development of a sensitive LC-MS/MS for the accurate, precise and reliable measurement of PF-5190547 and its hydroxy metabolite in human plasma. This study demonstrates the suitability of tacrine as IS for accurate quantification of PF-5190547 and hydroxy metabolite. This method demonstrated adequate reliability and reproducibility of the results within the analytical range. We have shown that PF-5190547 and hydroxy metabolite are stable in human plasma under different storage conditions. This assay was successfully employed for determination concentration of PF-5190457 and the metabolite in a phase 1b clinical trial.

Conflict of interest
Dr. Obach is an employee of Pfizer Pharmaceuticals. Other authors declare no potential conflict of interest with any commercial entity whose products were used in the study.
Tables Legend Table 1. Mass spectrometry parameters for analytes and IS Table 2. Summary of calibration standards from six validation runs Table 3. Summary of QC samples from six validation runs Table 4. Summary of stability studies

MANUSCRIPT-II
This manuscript has been prepared for submission to "Drug Metabolism and Disposition" The was identified as a potent and selective ghrelin receptor inverse agonist. The present study characterizes the routes of biotransformation of this compound in vivo and then further explores in vitro metabolism using subcellular fractions of human liver and primary hepatocytes. Following oral administration of PF-5190457, hydroxyl metabolites were observed, including one that had not been observed in human liver microsomal incubations. This major hydroxy metabolite was biosynthesized using liver cytosol and the site of hydroxylation was shown to be on the pyrimidine using nuclear magnetic resonance spectroscopy. The chemical inhibitors of AO (raloxifene) and XO (febuxostat) inhibited the formation of the major hydroxy metabolite in human liver cytosol. Both the enzymes were involved in the metabolism of the drug whereas, greater inhibition was observed to be achieved using AO inhibitor indicating AO to be a dominant enzyme. The Clint of the drug in human liver cytosol was 0.002 mL/min/mg protein. This study demonstrates the importance of considering the molybdenum-containing oxidases during the development of new drug entities.

Introduction
Ghrelin is a 28 amino acid peptide released from the stomach fundus and has a key role in growth and energy balance in the body (Korbonits et al., 2004). It is an endogenous ligand binding to the growth hormone secretagogue receptor in the hypothalamus resulting in increased food intake and decreased energy expenditure . Leptin has an opposite effect to that of ghrelin and both the hormones together regulate energy homeostasis in the body (Yildiz et al., 2004). Ghrelin is used as an orexigenic agent to stimulate food intake as shown in clinical studies (Nagaya et al., 2004;Neary et al., 2004;Nagaya et al., 2005). Moreover, intravenous ghrelin administration suppresses glucose dependent insulin secretion in healthy subjects (Tong et al., 2010). It has been observed that neurotransmitters and peptides regulating food intake can also influence craving for alcohol in alcoholic individuals (Leggio et al., 2011;Volkow et al., 2013). In a human study, intravenous administration of ghrelin (3 mcg/kg in 45 subjects) increased the craving towards alcohol in alcohol dependent heavy drinkers . Furthermore, in a rat study, it was demonstrated that antagonizing ghrelin receptor could prove as a potential target to treat alcohol addiction (Suchankova et al., 2013). Thus, inhibition of ghrelin and growth hormone secretagogue system can potentially lead to the discovery of novel pharmacological treatments for alcoholism. [3.5]nonan-7-yl)ethan-1one; Fig. 1) is a member of spiro-azetidino-piperidine series identified as a potent ghrelin receptor inverse agonist (Bhattacharya et al., 2014). It is presently undergoing clinical trials at the National Institutes of Health for the treatment of alcohol use disorder . In healthy volunteers, PF-5190457 was observed to be well tolerated  and it was found to be safe in a phase 1b study in subjects with alcohol use disorder (Lee et al, to be submitted for publication). The absorption of the drug after oral administration was rapid with a Tmax of 0.3-3 h and the elimination halflife (t1/2) was found to be ~8 h .
Limited information is available on the biotransformation of PF-5190457.
The assessment of metabolic pathways of a new drug is important as it helps to understand the pharmacokinetics and pharmacodynamics of the compound in human. It is essential to identify the enzymes responsible for biotransformation, as this information contributes to an understanding of interindividual variability in pharmacokinetics that can be due to genetic polymorphism or drug-drug interactions. In addition, metabolites generated by these enzymes could be pharmacologically active or have off-target effects (Lin and Lu, 1997;Kumar and Surapaneni, 2001).
The objective of the present study was to characterize the biotransformation of PF-5190457 in vitro and to identify the major human circulating metabolites. Specifically, the objectives include 1) to characterize the circulating metabolites in vivo in human plasma, 2) to determine of formation of metabolites in vitro, in sub-cellular fractions of human liver and human hepatocytes, 3) to confirm the structure of identified metabolites by mass spectrometry and NMR analyses 4) to assess the kinetic parameters of the metabolite formation, and 5) to identify the enzyme/s responsible for the formation of the major circulating metabolite using specific inhibitors for each enzyme.

Preparation of subcellular fractions
Human liver tissues (n=104) were obtained from Xenotech (Lenexa, KS, USA). Microsomal (HLM) and cytosolic (HLC) fractions were prepared by differential ultracentrifugation from human liver as described previously . Total protein concentration for the HLM and HLC were determined using Pierce BCA (bicinchoninic acid) protein assay kit (Thermo Fisher Scientific Inc, Waltham, MA, USA) and the fractions were stored in -80°C until further analysis.

HLM and HLC Incubations
The in vitro metabolism of PF-5190457 was studied using HLM and HLC.
PF-5190457 was incubated at relatively high concentrations (50-100 µM) with pooled HLM and HLC in a total volume of 50 µL of 100 mM potassium phosphate buffer (pH-7.4) such that the final concentration of methanol was less than 0.25%. Incubations were conducted at 37°C in an oscillating water bath. The total protein concentration used for both HLM and HLC ranged from 0.05 -2 mg/mL.
PF-5190457 was also incubated in only 100 mM potassium phosphate buffer (pH-7.4) without HLM or HLC as control. Microsomal incubations were conducted with and without the addition of NADPH. The reaction was terminated by addition of 200 µL of ice-cold methanol at various incubation times (0-120 min).
Subsequently, the samples were centrifuged for 5 min at 10,000 rpm to remove the precipitated protein.

LC-MS/MS analysis of HLM and HLC samples
The supernatant obtained from the samples was analyzed using an Acquity UPLC coupled with a Xevo-TQ mass spectrometer (Waters Corp, Milford, MA, USA). Samples (5 μL

Incubations in liver cytosol of various species
PF-5190457 was incubated at 20 µM, with pooled mouse (male and female), rat and monkey (male and female) liver cytosol (2 mg/mL) in a total volume of 1 mL potassium phosphate buffer (100 mM; pH 7.5). Incubations were performed at 37°C for 75 min and then stopped by addition of acetonitrile (5 mL).
Samples were centrifuged at 1700 g for 5 min and the supernatant was evaporated in vacuo, reconstituted in 1% formic acid (0.2 mL) for analysis by HPLC-UV-MS.
The incubation was terminated with addition of CH3CN (40 mL) and the precipitated material was removed by spinning in a centrifuge for 5 min at 1,700 g. The supernatant was partially evaporated in a Genevac vacuum centrifuge for 2 hr. To the remaining mixture was added 0.25 mL neat formic acid, 0.25 mL CH3CN, and water to a final volume of 25 mL. This mixture was spun in a centrifuge at 40,000 g for 30 min. The clarified supernatant was applied to an HPLC column (Polaris C18, 4.6 x 250 mm; 5µ particle size) through a Jasco HPLC pump at a rate of 0.8 mL/min. After the entire sample was applied, an additional ~5 mL of mobile phase (0.1% formic acid containing 1% CH3CN) was pumped through the system. The column was moved to a Thermo LTQ HPLC-MS system containing a photodiode array detector and a mobile phase gradient was applied to elute material of interest. The mobile phase was comprised of 0.1% formic acid in water and CH3CN and was run at a flow rate of 0.8 mL/min.
The gradient began at a composition of 2%B for 5 min followed by a linear gradient to 25%B at 70 min. The eluent passed through the photodiode array detector scanning from 200-400 nm and then to a splitter (ratio was

Enzyme Kinetic Studies
The formation of major hydroxy metabolite in rAOX and HLC was studied to determine the enzyme kinetic parameters. Before the assessment of the enzyme kinetics, protein concentration and incubation time for the kinetic studies were evaluated to ensure the linearity of major hydroxy metabolite formation with respect to protein concentration and time. It was also seen that substrate utilization during the reaction was less than 10%.

rAOX incubations
PF-5190457 (0.5-125 µM) was incubated with rAOX (0.090 mg/mL) in 100 mM potassium phosphate buffer. The incubations were conducted at 37°C in shaking water bath. The incubation mixture volume was 50 µL and all incubation samples were in triplicate. The reactions were terminated after 120 min using icecold methanol containing 2.5 ng/mL tacrine as IS.

Screening of allopurinol and oxypurinol in HLC.
The human livers utilized in this study were obtained from Sekisui XenoTech LLC which is typically retrieved from brain dead individuals for organ transplantation purpose but could not be transplanted. As such, the livers were perfused with the University of Wisconsin Cold Storage Solution (UW) containing allopurinol. However, it is known that allopurinol and its primary metabolite oxypurinol to be potent inhibitors of xanthine oxidase (XO) activity .
To detect the presence the XO inhibitors, HLC (n=104) were tested for the presence of allopurinol and oxypurinol using the same LC-MS/MS instrumentation and mobile phases as described earlier in HLC and HLM analysis. The flow rate was 0.3 mL/min, the column was equilibrated at initial conditions of 90% solvent A for 1.2 min and then to 10% at 2 min. Over the next minute, solvent A was increased to 90% and then kept constant for a re-

LC-MS/MS analysis for kinetic studies
All the samples from rAOX and HLC were centrifuged for 5 min at 10,000 rpm. The supernatants were removed and analyzed.

Incubation of PF-5190457 in HLC (with AO and XO inhibitors)
PF-5190457 (25 µM The reactions were terminated at 5 min using methanol containing 2.5 ng/mL tacrine as internal standard (IS). The samples were analyzed using LC-MS/MS as described under enzyme kinetic studies.

Data Analysis
Estimations of the maximum rate of major hydroxy metabolite formation The apparent in vitro intrinsic clearance (Clint) was calculated as Vmax/Km.

Characterization of PF-5190457 Metabolites in Human Plasma
The circulating metabolites of PF-5190457 were determined in the plasma of human subjects who were administered PF-5190457 orally in a phase 1b clinical study (Fig. 2 Fragmentation of this ion produced fragments at m/z 225 and 351 (Fig. 4). This metabolite was not observed in HLM supplemented with NADPH.

Hepatocytes
The formation of hydroxyl metabolites of PF-5190457 was observed in human hepatocytes as shown in the chromatogram (Fig. 5). The traces were extracted ion chromatograms of m/z 529.2376 (5 ppm tolerance) representing the protonated molecular ion of a hydroxylated metabolite. The addition of ABT, a broad-spectrum P450 inactivator inhibited the formation of the apparent minor metabolites at Rt 4.12,4.62,and 5.57 min and had no effect on the metabolite eluting at Rt 3.98 min suggesting that CYP mediated metabolism for the minor metabolites and not the major metabolite. It was also observed that addition of hydralazine inhibited formation of the metabolite at Rt 3.97 min indicating that AO could be responsible for the generation of this metabolite.

Incubations in liver cytosol from various species
The formation of hydroxy metabolite (513+13) was observed in cytosolic incubations of PF-5190457 in mouse (male and female), rat (male), monkey (female) indicating the metabolism to be independent of NADPH (Fig. 6). The hydroxy metabolite formed in cytosol of various species had similar retention time to that of the metabolite formed in HLC.

NMR Analysis
Characterization of biosynthesized hydroxy metabolite by NMR analysis demonstrated that the site of hydroxylation was on the carbon between the nitrogen on the 4-methylpyrimidine moiety (Fig. 7). 1 H COSY, HMQC, and HMBC spectra were obtained to support the hydroxylation of pyrimidine moiety (Supplemental data Fig. 2, 3 (a) and (b)).

Enzyme Kinetic Analysis
The HLC that were free (n=9) of allopurinol and oxypurinol were pooled together for further kinetic and inhibition experiments. The kinetics for the formation of major hydroxy metabolite from PF-5190457 was determined by incubating in rAOX and HLC. Representative Michaelis-Menten kinetic plots in rAOX and HLC are depicted in Fig. 8 and Fig. 9. Substrate inhibition kinetics at higher concentration of substrate was observed for rAOX incubations. The Km and Vmax values for the formation of hydroxy metabolite in rAOX and HLC were found to be 6.9 μM and 0.35 nmol/min/mg protein and 42 μM and 0.12 nmol/min/mg protein respectively. The Clint in rAOX and HLC were 0.05 and 0.002 mL/min/mg protein respectively.

Biotransformation of PF-5190457 is mediated by AO and XO
The effect of AO (raloxifene) and XO (febuxostat) inhibitors on the metabolism of PF-5190457 was investigated. There was considerable (>75%) but not complete inhibition of formation of the major hydroxy metabolite with 50 µM raloxifene (Fig. 10a). The extent of inhibition with febuxostat was greater than 50% at 1 µM (Fig. 10b) and further increases in inhibitor concentration did not inhibit the formation of metabolite.

Discussion
Characterization of circulating metabolites from patients orally administered with PF-5190457 revealed the biotransformation pathways for this ghrelin receptor inverse agonist. The primary route of metabolism observed in humans was hydroxylation yielding a major hydroxy metabolite of PF-5190457).
In vitro experiments showed that this metabolite was generated in human liver  (Krenitsky et al., 1972;. In contrast to electrophillic oxidation by CYP450 enzymes, both AO and XO enzymes display nucleophilic attack on carbon atom which is electron deficient or carbon adjacent to N-heterocycle (eg. purines, pyrimidines and pteridines) resulting in the formation of distinct metabolites (Krenitsky et al., 1972). Furthermore, the use of inhibitors indicated contribution of both AO and XO enzymes to the formation of hydoxy pyrimidine metabolite of PF-5190457.
The data also suggest AO as a dominant enzyme in comparison to XO in the biotransformation of PF-5190457.
AO and XO enzymes belong to a family of molybdo-flavoenzymes and have a high level of similarity between their amino acid sequence homology Terao et al., 2000;. XO can exists as two interconvertible forms as xanthine oxidase and xanthine dehydrogenase unlike AO (Della Corte et al., 1969). AO was found to be one of the highly abundant enzymes in HLC and the expression of AO was greater in comparison to XO in human liver (Nishimura and Naito, 2006;Sodhi et al., 2015). Physiological relevance of AO is still not known but XO plays an important role in catalyzing hypoxanthine to xanthine and later to uric acid   (Weidert et al., 2014).
Allopurinol has been widely used as a known XO inhibitor but recently it was identified that febuxostat is a more potent inhibitor of XO than allopurinol (Malik et al., 2011). There are not any important clinical drug-drug interactions associated with inhibition of AO but as growth in new chemical entities metabolized by AO is increasing, there could arise a need to characterize in drug-drug interaction potential at AO level (Obach, 2004;. It is a challenge to capture the drug interactions associated with these enzymes especially for AO enzyme in animal models as there exist species differences for expression and activity along with disparity with gender and strain (Itoh, 2009;Crouch et al., 2017;Mao et al., 2017).
Both potent XO inhibitors, allopurinol and febuxostat are clinically used for treatment of chronic gout and hyperuricemia. This condition is also present in disease states like metabolic syndrome, diabetes, chronic liver disease and cardiovascular disease (Hu and Tomlinson, 2008;George and Minter, 2017) some of which are target population for PF-5190457. There have been drug-drug interactions reported in man through the inhibition of XO substrates. For example, in one clinical study, allopurinol significantly inhibited 6-mercaptopurine which is a substrate for XO when administered orally and there was significant increase in peak plasma concentration and area under the curve of 6mercaptopurine in patients with acute lymphoblastic leukemia (Zimm et al., 1983).

Inter-individual variability between donors has been observed in in vitro
studies for hepatic AO and XO activity (Guerciolini et al., 1991;. However, single nucleotide polymorphisms in the AO and XO genes has not been extensively studied in various populations (Levartovsky et al., 2000;Gok et al., 2003;. The pharmacokinetic profile of the PF-5190457 and its major hydroxy metabolite has been investigated in plasma samples obtained from human subjects during the phase 1b study which is described elsewhere (Manuscript IV). It is important to identify the other metabolic pathways and route of elimination of PF-5190457 in the context of a mass balance study.

Conflict of interest
Dr. Obach is an employee of Pfizer Pharmaceuticals. Other authors declare no potential conflict of interest with any commercial entity whose products were used in the study.  Tables   Table II-          were sensitive to ABT. The peak eluting at 3.97 min (denoted with red arrows) was insensitive to ABT but was inhibited by hydralazine.

Introduction
Aldehyde oxidase (AO), along with xanthine oxidase (XO), belongs to the family of molybdo-flavoenzyme or molybdenum co factor containing enzymes and does not require NADPH as a cofactor for oxidation .
Unlike XO which catalyzes uric acid, the physiological function of AO in not known yet . In the recent year, AO is gaining significance in drug discovery programs because of a consorted effort to reduce the contribution of cytochrome P450 (CYP450) mediated biotransformation to the elimination of new drugs, Rashidi and Soltani, 2017). The new drug molecules are designed to have nitrogens in their rings like purines pyrimidines and pteridines to enable aldehyde oxidase metabolism (Lepri et al., 2017;Paragas et al., 2017). In contrast to electrophilic attack by CYP450 enzymes, aldehyde oxidase follows nucleophilic attack using the oxygen from the water Garattini et al., 2003). Various isoforms AOX1, AOX3, AOX3L1, and AOX4, have been identified with species differences. The functionally active aldehyde oxidase that is expressed in the human liver cytosol as AOX1 along with two other pseudogenes AOX3 and AOX3L1 (Kurosaki et al., 2013).
While a limited number of studies are published assessing the AOX1 levels in HLC, it is known that a significant degree of variability is associated with the expression of this enzyme (Barr et al., 2013;Fu et al., 2013). In one of the studies, the AOX1 levels reported in three batches of pooled HLC was found to be between 2 -40 pmol/mg (20 fold) (Barr et al., 2013). The other study assessed the levels in 20 individual HLC donors and the results observed were much tighter ranging from 0.74 -2.30 pmol/mg (~3 fold) (Fu et al., 2013). The variation between both the studies was observed to be about 17 -28-fold. Both the studies mentioned above used the absolute method of quantification (MRM) using mass spectroscopy.
The use of mass spectrometry is gaining popularity over traditional methods, such as Western blotting, enzyme-linked immunosorbent assays (ELISAs) and reverse-transcriptase polymerase chain reactions (RT-PCR) to monitor or quantify the expression of proteins (Han et al., 2008;Aebersold et al., 2013). Label-Free relative quantification approach for studying protein abundance of metabolizing enzymes in biological samples using mass spectrometry acquisition methods like data independent acquisition (DIA) and data dependent acquisition (DDA) (or information dependent acquisition) have been explored in the recent years (Hopfgartner et al., 2012). Our lab has recently published on using sequential windowed acquisition of all theoretical fragment ion mass spectra (SWATH-MS-MS), which is a DIA method, to study relative quantification of enzymes using a label free approach .
In the present study, we used the DDA, and SWATH-MS to quantify the

Human liver bank.
Human liver tissue (N=104) was purchased from Sekisui XenoTech LLC (Kansas City, KS). The livers were well characterized for age, gender, ethnicity, the cause of death, cold ischemia time, liver and body weight, smoking, alcohol consumption, diabetes and NAFLD. The information was available for ≥95% of the samples.

HLC Preparation and Digestion.
Preparation of HLC and digestion of the samples was performed as described previously . Briefly, human liver samples were homogenized using a bead homogenizer (Bead Ruptor 24, Omni International, Kennesaw, GA). Ultracentrifugation was done to separate microsomes and cytosol. The cytosolic fraction thus obtained was stored in -80°c for future use.
Cytosolic protein concentration was estimated using a bicinchoninic acid method

LC-QTOF/MS Analysis.
All the cytosolic digested samples were analyzed on a SCIEX 5600 TripleTOF mass spectrometer equipped with a DuoSpray ion source (SCIEX, Concord, Canada) coupled to Acquity UHPLC HClass system (Waters Corp., Milford, MA, USA) which has a with a binary pump and a built-in column heater.
Mass spectral analysis was carried out using a positive electrospray ionization mode. Chromatography separation of peptides was accomplished within 11.30 min using gradient elution method using an Acquity UHPLC Peptide BEH C18 (2.1 × 150 mm2, 300 Å, 1.7 μm) analytical column. An Acquity VanGuard precolumn (2.1 × 5 mm 2, 300 Å, 1.7 μm) was used prior to the analytical column. The analytical column and autosampler temperature consisting digested samples were kept at 40 °C and 10°C, respectively. Mobile phase A consisted of 98% water, 2% acetonitrile containing 0.1% formic acid and mobile phase B 98% acetonitrile, 2% water containing 0.1% formic acid. The mobile phase was delivered at 100 μL/min flow rate with a gradient method and total run time of 60 min. The initial conditions of mobile phase composition are as follows: (98% A) was maintained for 3 min, 60% to 90% A from 3 to 48 min, and 20% A held till 52 min and returned to initial conditions for equilibration. The amount of protein per injection on the column was 10 μg. In each batch, Trypsin-predigested betagalactosidase peptides were injected (∼30 pmol/injection) as quality control every 10 samples during the analysis to monitor mass calibration of the TOF detector and normalization of intensity during relative quantification of AOX1 protein (described below). The average intensity of the β-galactosidase peptide (APLDNDIGVSEATR) in a batch was used for data normalization of the respective batch of samples.

DDA and SWATH-MS Data Acquisition Settings.
Two acquisition methods, DDA and SWATH-MS method were used to identify and quantify AOX1 protein. Analyst TF 1.7 was used to acquire data during the study (SCIEX, Framingham, MA). The detailed settings for source and compound parameters for both the acquisition methods are as described in previously published paper from our lab .

Generation of spectral reference library
The data files of HLC samples obtained from the DDA acquisition were analyzed using the database search engine ProteinPilot 5.0 (SCIEX; Framingham, MA, USA) The data were searched against the UniProt protein database. The search was specified using the following parameters in the Paragon method: sample type -identification; cysteine alkylation -MMTS; digestion -Trypsin; instrument -TripleTOF 5600, species -Homo sapiens; search effort -thorough and false discovery analysis -none.

Method 1 Based on DIA-SWATH-MS MS and Skyline analysis.
The spectral library generated from DDA files was uploaded in Skyline to identify AOX1 protein, and SWATH-MS data files were processed using the full scan MS/MS filtering at a resolution of 10000. Three unique AOX1 peptides were identified and curated for reproducible fragment ions. The total area of representative peptides for AOX1 protein was summed, and resulting intensity was normalized by total intensity of tryptic peptide of β-galactosidase. MultiQuant v 3.0 (SCIEX, Framingham, MA) was used to retrieve intensity for APLDNDIGVSEATR peptide and was subsequently used for normalization among different batches as described above (Nakamura et al., 2016).

Method 2 Based on DDA acquisition and MaxQuant analysis.
The raw data files obtained from DDA data was analyzed for AOX1 using MaxQuant (Version 1.5) and was searched against UniProt human protein database at a false discovery rate (FDR) of 0.01. The absolute protein concentration was calculated from label-free quantification (LFQ) intensities using "Total Protein Approach" (Wisniewski and Rakus, 2014).

Statistical analysis.
Statistical analysis was performed with SPSS version 24 (IBM Analytics, Armonk, NY), and Prism ® version 6 (GraphPad Software Inc., San Diego, CA) was used for graphs. Non-parametric Mann-Whitney U test and Kruskal-Wallis test (2-tailed) were used to study the effect of three or more groups. The correlation was studied using nonparametric Spearman correlation analysis. P <0.05 was considered significant for all the statistical tests and correlation analysis.

SWATH-MS based relative AOX1 levels
Three unique peptides for AOX1 protein DILADVVQK, MIQVVSR, GTSTETVPNANISGGSVVADLNGLAVK and three fragment ions per peptide for the protein were identified. AOX1 levels from SWATH-MS files are represented after normalization Selected peptides for the relative quantification of the AOX1 protein were correlated with each other and also, with sum of the three peptides to validate the selection of peptides (Fig 1a and b). The correlation between the surrogate peptides (Spearman, r = 0.94, p<0.0001) was significant. Likewise, the correlation between one of the selected peptides (DILADVVQK) of AOX1 and sum of the peptides was significant (Spearman, r = 0.98, p<0.0001). Two other peptides had similar significant correlation (data not shown) generating confidence in the selection of peptides.

DDA based absolute AOX1 levels
The AOX1

Correlation between SWATH-MS and DDA based results
The correlation between the AOX1 levels measured via both the acquisition methods (DIA and DDA) was significant (Spearman, r = 0.72, p<0.0001) (Fig 3). The results indicate the reliability of either of the acquisition methods for determination of AOX1 levels.

Effect of demographic factors on AOX1 levels.
The effect of various factors like gender, age, weight, drug usage, ethnicity, smoking, alcohol consumption along with disease conditions like diabetes and NAFLD (non alcoholic fatty liver disease) was evaluated on AOX1 levels.

Cytosolic fatty acid synthase (FAS). FAS is observed to be a diagnostic marker
for progression in NAFLD (Dorn et al., 2010). As the disease progresses, the FAS expression is expected to increase as seen in Fig 4. This enzyme was used as a positive control to determine the quality of data generated from mass spectroscopy.

Effect of Ethnicity.
The number of African American, Caucasian and Hispanic included in this study are as shown in Table 1. In general, African Americans exhibited marginally higher AOX1 levels in comparison to Caucasian and Hispanic donors.

Effect of Alcohol.
Alcohol consumption in the liver donors were evaluated and scores were determined based on the drinking scale (None = 0, Safe = 1, Moderate = 2, Heavy = 3). The trend of AOX1 levels decreased with increase in alcohol consumption but not significant (P>0.4) (Fig 5) ( Table.2)

Effect of other demographic factors.
Other demographic factors such as gender, age, weight, drug usage, smoking, along with disease conditions like diabetes and NAFLD (non alcoholic fatty liver disease) were not detected to have an impact on AOX1 levels based on our results. Table. 2 shows the effect of selected demographic factors on AOX1 with mean ±SE

Discussion
The AOX1 levels measured in HLC were found to be variable in the present study. Cytosolic AOX1 levels in human livers were quantified using two different acquisition methods in mass spectroscopy. Our lab had initially developed a DIA method using SWATH-MS method to determine the relative protein levels of various drug metabolizing enzymes which was published recently . DDA was used for identification of proteins in the earlier method (details in the method section). In due course, MaxQuant a quantitative proteomics software was utilized which allows both in identifying and quantifying the proteins of interest using the raw files generated from DDA . Both the acquisition methods (DIA and DDA) demonstrated good correlation with each other and reliability in using either method to quantify AOX1 protein. The cytosolic FAS protein (a marker for progression of NAFLD) assessment gave confidence in the reliability of the quantified protein levels using mass spectroscopy. The leverage of total protein approach in MaxQuant is that it gives accurate estimations of protein concentrations . The absolute values of pmol of AOX1/mg of total cytosolic protein values generated from MaxQuant in the present study were comparable to other studies (Barr et al., 2013;Fu et al., 2013). The variation in our study was observed to be about 25-fold (2 -49 pmol/mg) in a novel bank of human livers. The AOX1 levels quantified were much closer to the results reported by Barr et al. in pooled HLC (Barr et al., 2013).
The expression of drug metabolizing enzymes can be influenced by many intrinsic and extrinsic factors Fisher et al., 2009;Yang et al., 2013;He et al., 2015). Most importantly factors like genetic polymorphism (Meyer, 2004), epigenetic modifications like DNA methylation, histone protein modification or interaction with microRNAs (Chuang and Jones, 2007;Ingelman-Sundberg and Gomez, 2010), and non genetic factors like age (Kinirons and O'Mahony, 2004), gender (Beierle et al., 1999) or disease state (Elbekai et al., 2004) can influence the expression of metabolizing enzymes. Pharmacokinetic variability in individuals can be largely explained by the above mentioned factors influencing the metabolizing enzymes and the knowledge of these factors is important to monitor the drug response (Lin, 2007).
AOX1 is one of the highly abundant enzymes found in human liver Nishimura and Naito, 2006). Non alcoholic fatty liver disease (NAFLD) has demonstrated to influence the AOX1 levels. NAFLD is characterized by a spectrum of conditions ranging from simple steatosis, nonalcoholic steatohepatitis with fibrosis progressing to cirrhosis (Targher et al., 2008;Younossi et al., 2011). The rats fed with high-fat diet and primary hepatocytes loaded with palmitic acid showed an increase in AOX1 expression (Neumeier et al., 2006). It was observed that the primary hepatocytes for the study were isolated and cultured from tissue samples obtained were not from healthy subjects but from patients with liver tumors of colorectal cancer who were undergoing hepatectomy. The levels of AOX1 could also be influenced by disease conditions like cancer. Certain phase I and II drug metabolizing enzymes, AOX1, CYP1B1, GSTM3 and GSTP1 gene and protein expression were found to be elevated in brain tumor specimens obtained from patients (n=77) undergoing primary resection (Stavrinou et al., 2015). Whereas, in the present study NAFLD did not have a significant impact on the AOX1 levels.
In an in vitro study (Fu et al., 2013) poor correlation was observed between AOX1 activity and expression levels in HLC. Several demographic factors along with protein mis-folding, polymorphism and co-factor deficiency were suspected to be the reasons for observed poor correlation, but no clear conclusions were drawn owing to their small sample size (n=20). In the same study, chronic alcohol consumption in liver donors (n=2) decreased the AO activity significantly but not the AOX1 levels. The AOX1 levels in the present study were not altered significantly with increase in alcohol consumption. In other in vitro studies, considerable inter-individual variability in the AO metabolism has been observed using different substrates . It would have been beneficial if the studies had AOX1 levels measured to correlate with the observed activity and if any other demographic factors or posttransitional modifications are influencing the activity of the protein.
AO contains molybedenum and iron sulfur centers along with flavin adenine dinucleotide . Iron deficiency diet decreased the activity of AO in rats and increasing the iron reverted the activity to normal indicating iron to have a significant impact on AO enzyme .
It is known that XO belongs to the same family as AO and is closely related enzyme to AO . In a rat study, XO levels and activity were observed to be decreased with iron-depleted diet and increased when supplemented with iron . Both these studies indicate iron could play an important role in modulating the AOX1 levels.
Many other unknown factors could be the reason for observed variability in the AOX1 levels in our study. Recognizing the sources of variability could be challenging as there may be other underlying unknown epigenetic factors, disease state or dietary constituents influencing the enzyme. To the best of our knowledge, for the first time, the effect of demographic factors on AOX1 levels in a large number of human livers was studied.

Conflict of interest
Dr. Obach is an employee of Pfizer Pharmaceuticals. Other authors declare no potential conflict of interest with any commercial entity whose products were used in the study. Tables   Table. III  Legend of Figures   Fig. 1. Correlation plots   Correlation between two acquisition methods SWATH-MS and traditional DDA was considered significant (p<0.05) using correlation coefficient Spearman r    Column and error bars represent mean±SE. P >0.4 reported from nonparametric Kruskal-Wallis test (2-sided). protein levels were obtained by mass spectroscopy using a label free quantitation approach and information dependent acquisition.

RESULTS:
The half-life of parent and metabolite was found to be approximately 6 and 13 h respectively for both doses. The CLint in HLC for PF-5190457 ranged from 0.4 -7.1 µL/min/mg protein with 17-fold variability. The estimated AOX1 protein expression were much closer ranging from 16.8 -44.1 pmol/mg (2.6-fold variability). The correlation between AOX1 protein expression and aldehyde oxidase (AO) activity was poor.

CONCLUSION:
The pharmacokinetics of PF-5190457 and its major metabolite was evaluated after oral administration of 50 and 100 mg and importantly the half-life of the metabolite was found to be longer than the parent. Inter-individual variability in the hydroxy metabolite formation was observed in HLC indicating a Introduction PF-5190457 is a ghrelin receptor inverse agonist and is currently undergoing clinical trials to treat alcoholism and type 2 diabetes mellitus [1][2][3]. This drug was originally developed by Pfizer and the tolerability of the compound in healthy volunteers was evaluated. The pharmacokinetics of the drug in healthy volunteers (n=35) was determined at single (2, 10, 40, 50, 100, 150, and 300 mg) and multiple (2,10,40 and 100 mg BID for two weeks). The absorption of the PF-5190457 was observed to be rapid and the elimination half-life was found to be ~9 h [4].
We have previously showed that PF-5190457 is metabolized to a hydroxy metabolite by the molybdenum co-factor containing enzymes AO and xanthine oxidase (XO). The contribution of AO in the metabolism of the drug was identified to more dominant (manuscript II). AO enzyme is gaining increasing importance in the metabolism and clearance of new drug entities in the recent years [5,6].
There are various in vitro studies in S9 fraction, cytosol and hepatocytes demonstrating high interindividual variability in AO activity with use of various substrates [7][8][9]. For example, AO activity measured in cytosol from same donors using carbazeran, zoniporide and phthalazine as probe substrates varied 90, 42 and 17-fold respectively [10].
In the present paper, we determine the concentration of PF-5190457 and its hydroxy metabolite in alcoholic patients (N=12) enrolled in a phase 1b clinical study in subjects with alcohol use disorder. In an effort to address the variability of PF-5190457 metabolism with respect to AO activity, the metabolite formation was measured in HLC from 10 individual donors. In addition, the measured activity was correlated with the expression of AO protein and the characteristics of HLC donors were examined to understand the impact of demographic variables such as age, gender, ethnicity, smoking, alcohol, medical history.

Pharmacokinetic study in Alcohol Dependent Subjects
Phase 1b clinical study of PF-5190457 was conducted in non treatment seeking alcohol dependent individuals at NIH clinical center at Bethesda. Men and women between 21 -65 years were eligible. The study was approved by the University of Rhode Island Institutional Review Board HU# 123.172. The participants of the study have filled out the informed consent form prior to the study. Briefly, the key inclusion criteria were heavy drinking (on an average for men -21 drinks and women -14 drinks a week) and good health which was confirmed by medical history, physical examination, ECG, blood/urine lab tests.
Moreover, the female subjects must be of non-childbearing potential, and male subjects must follow contraception for a period of 28 days after dosing. Briefly the key exclusion criteria were interest in receiving treatment for alcohol use disorder, diagnosis of substance dependence (other than alcohol and/or nicotine), schizophrenia, bipolar disorder, or other psychoses, clinically significant medical abnormalities, heart rate >100 at screening on two separate measurements given potential of study medication to increase heart rate and BMI less than or equal to 18.5.
This phase 1b study was a within-subject design and a single blind doseescalating placebo controlled inpatient study. Subjects (N=12) were admited to the clinical research center (NIH clinical center at Bethesda) for three visits (placebo, a dose of 50 mg and a dose of 100 mg). The subjects received PF-5190457 orally as solution every twelve hours for a total of 5 doses. Blood samples for pharmacokinetics analysis were collected in the morning of the third day of hospital stay after the administration of a fifth dose of PF-5190457 at various time points (0,15,30,45,60,90,120,150,180,210,240,360,480,1350 and 1440 min). The concentration of PF-5190457 and its hydroxy metabolite (PF-6870961) in plasma were measured using an LC-MS/MS assay specifically developed and validated for the purpose of this study as described in manuscript I. Briefly, the plasma extraction method involved simple precipitation in methanol and samples were analyzed using an Acquity UPLC system coupled with Xevo TQ MS detector (Waters Corp, Milford, MA, USA). The pharmacokinetic parameters were determined using non-compartmental analysis. Cmax (maximum observed drug concentration during the study), AUC0-24 is area under the plasma concentration-time curve measured for 24 h, using the trapezoidal rule, AUClast-∞ is based on the last observed concentration extrapolated to infinity time (Cplast/k), k is rate constant, AUC0-∞ is area under the curve from time 0 extrapolated to infinite time (AUC0-24 + AUClast-∞), Tmax is time to observe maximum drug concentration, and T1/2 is half-life determined by 0.693/K). The apparent volume of distribution and clearance were calculated using the relations Vz/F= Dose / (AUC0-∞* k) and Cl/F = Dose / AUC0-∞ respectively [11]. The pharmacokinetic parameters were evaluated for each subject at respective doses (50 and 100 mg).

Human Liver Selection for In vitro studies
The livers (n=10) for the in vitro study were selected from the existing novel human liver bank (n=104) in the lab purchased from Sekisui XenoTech LLC (Kansas City, KS). The primary criteria for selecting the livers in this study were based on ethnicity (African American -AA, Caucasian -CC). It is a common practice to perfuse livers with University of Wisconsin solution containing allopurinol (a well-known XO inhibitor) and its primary metabolite oxypurinol and they have shown to completely inhibit XO activity [12]. From our liver bank, we identified that only n=9 out of 104 were allopurinol free and none of the donors were African American. For the purpose of uniformity, the livers selected for the present study were containing allopurinol/oxypurinol which was detected by mass spectrometry analysis as described in Manuscript II. The livers were well characterized and documented for demographic factors like age, gender, BMI, weight of body and liver, ethnicity, alcohol consumption, smoking, along with medical history for diabetes and non-alcoholic fatty liver disease. Table 1 shows demographic characteristics of donors.

Human liver cytosol Incubations
The procedure that was used to prepare liver cytosol has been described elsewhere [13].

MS/MS
AOX1 levels (pmol/mg) in cytosolic fractions were determined using a label free approach on mass spectrometry as described in Manuscript III. Briefly, the digestion of proteins in cytosolic samples was performed using trypsin and the resulting peptides were analyzed using information dependent acquisition (IDA) mode on an AB Sciex TripleTOF 5600+ mass spectrometer. The subsequent data analysis was performed using MaxQuant (Version 1.5) proteomics software for protein (AOX1) identification by searching against UniProt human protein database (updated Oct 2016) and quantification. The maximum false peptide discovery rate was specified as 0.01. AOX1 protein was calculated based on the raw spectral protein intensity from label-free quantification (LFQ) intensities of the MaxQuant software output using the "Total Protein Approach" (TPA).

Statistical analysis
Statistical analysis was performed with SPSS version 24 (IBM Analytics, Armonk, NY), and Prism ® version 6 (GraphPad Software Inc., San Diego, CA) was used for graphing and calculation of kinetic parameters (Vmax and Km). Since most of the datasets were not normally distributed, nonparametric methods were generally used for statistical analyses. Non-parametric Mann-Whitney U test (2tailed) was used to compare the effect of ethnicity and other demographic factors on enzyme activity and AOX1 levels. The correlation was studied using nonparametric Spearman correlation analysis.

Pharmacokinetics analysis
The validated LC-MS/MS method (Manuscript I) has been successfully applied to measure the concentrations of PF-5190547 and its hydroxy metabolite after oral administration of the drug at 50 mg and 100 mg in human plasma. The mean concentration versus time profiles of the analytes in human plasma are shown in Figure 1 a and b. The plasma concentration ranges of the analytes were measurable within their calibration curves. The concentrations of PF-5190547 and its hydroxy metabolite ranged from 2.8 to 213.3 ng/mL and 3.9 to 73.8 ng/mL for 50 mg dose and 3.5 to 813.7 ng/mL, and 6.6 to 124.8 ng/mL for 100 mg dose, respectively. Pharmacokinetic details for the area under the curve from initial time to infinity (AUC0-∞), Cmax, Tmax, Vz/F, Cl/F and t1/2 are shown in Table 2 for both 50 and 100 mg doses. The half-life for the metabolite (14.7 ± 2.1 and 12.8 ± 2.2 h) at both the doses (50 and 100 mg) was observed to be longer in comparison to the parent (6.2 ± 1 and 5.8 ± 0.5 h).

Individual PF-5190457 metabolism in HLC
Formation of hydroxy metabolite of PF-5190457 in this study was mediated by AO as HLCs in the selected livers were devoid of XO activity. In general, individual variation in AO activity (Figure 2) was observed in all kinetic parameters for the hydroxy metabolite formation (Table 3). Km and Vmax values ranged from 8.1 to 68.3 µM and 0.02 to 0.41 nmol/min/mg protein respectively.
CLint was found to be between 0.4 -7.1 µL/min/mg (~ 17 fold variability). African American donors exhibited marginally lower AO activity in comparison to Caucasian donors. However, no significant association (P>0.1,) was observed between ethnicity and AO activity (Figure 3a). The effect of other demographic factors was examined and no significant relationship was observed.
Nevertheless, the effect of smoking on AO activity was observed to be significant with p<0.017 (Figure 4a). The limitation of this study was small sample size as there were only n=3 out of 10 livers were obtained from non-smoker individuals.

Expression of AOX1 protein in HLC and correlation to In vitro AO activity
The AOX1 levels obtained from data analysis using MaxQuant TPA approach  Table 1 did not significantly influence AOX1 levels. The AO activity characterized by rate of formation of hydroxy metabolite showed a weak correlation ( Figure 6) (Spearman r = 0.2 and p>0.5) with AOX1 protein levels.

Discussion
The in vivo study evaluated pharmacokinetics of PF-5190457 and its hydroxy metabolite in alcohol dependent subjects. The pharmacodynamics of the drug in the same study population is evaluated by our lab and will be discussed elsewhere.
Previously, published study determined PF-5190457 pharmacokinetics in healthy subjects [4] and the half-life of the parent was observed to be higher in comparison to our study in alcoholic subjects (~9 vs ~6 hours) whereas, the other parameters (Cmax, Tmax) were comparable. It could be because of differences in blood sampling points or differences in the population (healthy vs alcoholic subjects).
The kinetics of major hydroxyl metabolite was not reported previously. The findings from the present study also indicate that the half-life of the hydroxy metabolite in heavy drinkers was higher in comparison to the parent drug. The higher exposure of the metabolite in the body could be of an advantage if the metabolite would have any pharmacological significance and further evaluations with respect to safety profile and byproducts of the hydroxy metabolite are necessary. The clinical study for PF-5190457 recruited n=12 subjects and the majority of the subjects where African American (n=11). The differences in ethnicity could affect the pharmacokinetics and subsequent pharmacodynamics of the drug [14,15]. In addition, the variability of pharmacokinetic parameters is largely explained by genetic variation in metabolizing enzymes [16], disease state, physiological and environmental factors [17,18]. The primary objective for the in vitro study in the present paper was to determine differences in metabolism of PF-5190457 with respect to ethnicity (African American vs. Caucasians) and inter-individual variability.
PF-5190457 is predominantly metabolized by AO enzyme (Manuscript II). Wide variability has been reported for substrates metabolized by AO and the metabolism has demonstrated to be highly substrate dependent with high donor to donor variability [10,19,20]. An example, the AO activity was measured using different probe substrates N-[(2-dimethylamino) ethyl] acridine-4-carboxamide (DACA) and benzaldehyde in the same HLC donors (n=13). The CLint with respect to DACA and benzaldehyde was 18-fold and 5-fold respectively across the individual donors [21]. In our study, the CLint varied 17-fold for PF-5190457 across individual HLC. Further, weak correlation between the AOX1 protein levels and AO activity was observed which were in accordance with findings from another in vitro study in HLC [10]. Few studies published to determine correlation between AO protein and activity. The hydroxy metabolite formation was found to be reduced in AA but the non-significant change could be because of the small sample size of HLC. There is only one study so far, showing the influence of ethnicity on AO activity in cryopreserved hepatocytes using O6-benzylguanine as a probe substrate which showed no significant effect between AA (n=7) vs CC (n=50) on AO activity. The above-mentioned study too had several livers from AA donors. Further exploration of other demographic factors on PF-5190457 metabolism indicated smoking could lead to low AO activity but study in large sample size of donors would be essential for confident results. There is only limited information published on the impact of smoking on AO protein and activity. One study discussed no relationship between AO activity and smoking but it was based on 1 out of 13 HLC donors with tobacco use [21]. In one other study AO activity in large sample size (75 donors) was assessed using hepatocytes and the livers were characterized for tobacco use but the effect of smoking was not mentioned [22]. It is also known that nicotine is primarily metabolized by CYP1A2 to an iminium ion and then the second step of metabolism primarily involves AO enzyme [23][24][25]. Fu et al. showed chronic alcohol use in (n=2) HLC donors could influence the AO activity but the small sample size was a limitation. The effect of both smoking and alcohol use could have a significant impact on AO activity and warrants further examination in a larger number of samples.
AO protein is a homodimer and disruption in homodimer formation can affect AO activity [26]. Other reasons for variation in AO activity could be protein misfolding or genetic polymorphism [27]. Genetic polymorphism of hAOX1 was studied in Italian population and the presence of functionally inactive hAOX1 allelic variants enzymes was determined [26]. To the best of our knowledge there are no studies conducted exclusively in AA population to assess the presence of allelic variants.
One study determined the impact of common single nucleotide polymorphism on the AO activity in 75 donors in which there were only seven African American and 50 Caucasians and the rest were Hispanic. The study showed no statistical significance for polymorphism on AO activity [22].
Dietary constituents could also influence the AO activity [28]. AO enzyme contains iron along with molybdopterin cofactor and flavin adenine dinucleotide. It has been reported that iron deficiency could lead to lower AO activity as reported in a rat study. AO is the main enzyme for serotonin degradation and rats on iron deficient diet exhibited significantly increased concentrations of serotonin. The levels of serotonin reverted back to normal when the rats were supplemented with iron [29]. It is known that AO and XO enzyme share closely related structures. The XO expression and activity in lung were reduced in rats fed on an iron-depleted diet and increased in rats supplemented with iron [30]. AO belongs to the same family iron supplementation could be one of the reasons for the variability observed in AO activity.
The findings from the present study could be important as variability in PF-5190457 was observed across individual HLC. Since multiple factors could be contributing to variability in AO activity, larger sample size is necessary to address the connections.
In conclusion, the manuscript describes the successful application of LC-MS/MS method to a pharmacokinetic study of PF-5190547 at 50 and 100 mg in healthy volunteers in phase 1b clinical trials to treat alcoholism. The in vitro study shows donor variability in hydroxylation of PF-5190457 in human liver cytosol.

Conflict of interest
Dr. Obach is an employee of Pfizer Pharmaceuticals. Other authors declare no potential conflict of interest with any commercial entity whose products were used in the study.