Species differences in metabolism of EPZ015666, an oxetane-containing protein arginine methyltransferase-5 (PRMT5) inhibitor

Nathalie Rioux, Kenneth W. Duncan, Ronald J. Lantz, Xiusheng Miao, Elayne Chan-Penebre, Mikel P. Moyer, Michael J. Munchhof, Robert A. Copeland, Richard Chesworth & Nigel J. Waters

To cite this article: Nathalie Rioux, Kenneth W. Duncan, Ronald J. Lantz, Xiusheng Miao, Elayne Chan-Penebre, Mikel P. Moyer, Michael J. Munchhof, Robert A. Copeland, Richard Chesworth & Nigel J. Waters (2015): Species differences in metabolism of EPZ015666, an oxetane-containing protein arginine methyltransferase-5 (PRMT5) inhibitor, Xenobiotica
To link to this article: http://dx.doi.org/10.3109/00498254.2015.1072253

Published online: 31 Jul 2015.

Submit your article to this journal

Article views: 97
View related articles View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=ixen20

http://informahealthcare.com/xen

ISSN: 0049-8254 (print), 1366-5928 (electronic)

Xenobiotica, Early Online: 1–10
! 2015 Taylor & Francis. DOI: 10.3109/00498254.2015.1072253

RESEARCH ARTICLE

Species differences in metabolism of EPZ015666, an oxetane-containing protein arginine methyltransferase-5 (PRMT5) inhibitor

Nathalie Rioux1, Kenneth W. Duncan1, Ronald J. Lantz2, Xiusheng Miao2, Elayne Chan-Penebre1, Mikel P. Moyer1, Michael J. Munchhof1, Robert A. Copeland1, Richard Chesworth1, and Nigel J. Waters1

1Epizyme Inc., Cambridge, MA, USA and 2Quintiles Bioanalytical and ADME Labs, Indianapolis, IN, USA

Abstract
1. Metabolite profiling and identification studies were conducted to understand the cross- species differences in the metabolic clearance of EPZ015666, a first-in-class protein arginine methyltransferase-5 (PRMT5) inhibitor, with anti-proliferative effects in preclinical models of Mantle Cell Lymphoma. EPZ015666 exhibited low clearance in human, mouse and rat liver microsomes, in part by introduction of a 3-substituted oxetane ring on the molecule. In contrast, a higher clearance was observed in dog liver microsomes (DLM) that translated to a higher in vivo clearance in dog compared with rodent.
2. Structure elucidation via high resolution, accurate mass LC-MSn revealed that the prominent metabolites of EPZ015666 were present in hepatocytes from all species, with the highest turnover rate in dogs. M1 and M2 resulted from oxidative oxetane ring scission, whereas M3 resulted from loss of the oxetane ring via an N-dealkylation reaction.
3. The formation of M1 and M2 in DLM was significantly abrogated in the presence of the specific CYP2D inhibitor, quinidine, and to a lesser extent by the CYP3A inhibitor, ketoconazole, corroborating data from human recombinant isozymes.
4. Our data indicate a marked species difference in the metabolism of the PRMT5 inhibitor EPZ015666, with oxetane ring scission the predominant metabolic pathway in dog mediated largely by CYP2D.

Keywords
CYP2D6, CYP2D15, hepatocytes, interspecies scaling, metabolite identification

History
Received 5 June 2015
Revised 8 July 2015
Accepted 9 July 2015
Published online 3 August 2015

Introduction
Preclinical pharmacokinetic (PK) studies aim to characterize the absorption and disposition of new chemical entities in animals in order to support translational understanding of efficacy, safety and PK in man. The underlying species differences that can occur in the absorption, distribution, metabolism and excretion of xenobiotics then become of fundamental importance in the judicious selection of the most appropriate preclinical species for pharmacology and toxicology studies, as well as in improving preclinical extrapolation of PK to humans.
EPZ015666 is a first-in-class, orally available protein arginine methyltransferase-5 (PRMT5) inhibitor, that demon- strated potent cellular activity as measured by its ability to inhibit symmetric arginine dimethylation of SmD3, a PRMT5 substrate, in a time- and concentration-dependent manner. Treatment of Mantle Cell Lymphoma (MCL) cell lines with EPZ015666 led to inhibition of SmD3 methylation and cell killing, with IC50 values in the nanomolar range. Oral dosing

Address for correspondence: Nathalie Rioux, Ph.D., Principal Scientist, DMPK Epizyme, Inc., 400 Technology Square, 4th Floor, Cambridge, MA 02139, USA. Tel: 617 674 1792. Fax: 617 349 0707. E-mail:
[email protected]

of EPZ015666 demonstrated dose-dependent anti-tumor activity in multiple MCL mouse xenograft models (Chan- Penebre et al., 2015).
EPZ015666 exhibits low to moderate in vitro clearance in human, in part due to the introduction of an oxetane ring. Oxetanes have recently gained attention as attractive, stable and less lipophilic moieties for drug discovery (Burkhard et al., 2010, 2013; Wuitschik et al., 2006). A small number of drugs-containing di-substituted oxetanes, such as taxol, are well characterized, but very little is known about mono- substituted oxetane metabolism and in vivo clearance path- ways (Burkhard et al., 2013). The use of oxetanes was shown to improve solubility and human microsomal stability during optimization of a series of G-protein-coupled receptor 119 agonists (Scott et al., 2013). Stepan et al. (2011) recently illustrated the improvement in drug likeness of a series of arylsulfonamide molecules through the use of a 3-substituted oxetane motif, including increased metabolic stability in human liver microsomes (HLM), although these compounds had significantly higher clearance in rat liver microsomes.
The metabolism and PK of EPZ015666, a novel PRMT5 inhibitor, represent an interesting case study of the metabol- ism of a 3-substituted oxetane-containing drug candidate. The present work aimed to characterize the metabolism of

EPZ015666 and identify the metabolic pathways and enzym- ology contributing to the marked species differences in the PK of EPZ015666. This case study adds further insight of the potential mechanisms behind dog as an outlier in cross- species PK, underlying the importance of understanding interspecies metabolism and disposition in drug research.

Material and methods
Materials
EPZ015666 (N-[(2S)-2-hydroxy-3-(1,2,3,4-tetrahydroisoquinolin- 2-yl)propyl]-6-[(oxetan-3-yl)amino]pyrimidine-4-carboxamide) was synthesized by Epizyme (Cambridge, MA; Chan-Penebre et al., 2015). All other reagents were purchased from sources as described below.

Pharmacokinetic study in mouse, rat and dog
Studies were performed in accordance with the AAALAC International and NIH guidelines standards. Fasted male CD-1 mice (25–40 g; Vital River Laboratory Animal Technology Co. Ltd, China, n 3/group) and Sprague- Dawley (SD) rats (200–300 g; SLAC Laboratory Animal Co. Ltd., China; n 3/group) were treated with a single dose of EPZ015666, at 2 mg/kg, by tail vein injection (1 mg/mL in 20% N-N-dimethylacetamide in water for mice or in water, pH 7.0, for rats) or 10 mg/kg by oral gavage (2 mg/mL in 20% N-N-dimethylacetamide in water for mice or 0.5% methyl- cellulose for rats). Non-na¨ıve male Beagle-dogs (9.4–10.0 kg; Marshall Bioresources, China; n 3) were treated with a single dose of EPZ015666, at 2 mg/kg, via cephalic vein injection (1 mg/mL in water, pH 7.0). After a 7-day wash-out period, the same dogs were fasted overnight and administered EPZ015666 by oral gavage, at 10 mg/kg (2 mg/mL, 0.5% methylcellulose in water). Approximately 30 mL for mice or 100 mL of blood for rats were taken from animals via submandibular or retro-orbital bleeding for mice and jugular vein cannulation for rats at 0.0833, 0.25, 0.5, 1, 2, 8, 12 and
24 h post iv dosing and 0.25, 0.5, 1, 2, 4, 8, 12 and 24 h post oral dosing. Approximately 100 mL of blood was taken from dogs via a peripheral vein pre-dose and at 0.0833, 0.25, 0.5, 1,
3, 6, 9, 12 and 24 h post iv and po dosing. Blood samples were transferred into K2-EDTA tubes and placed on wet ice prior to centrifugation at 4 ◦C (7000 rpm, 5 min) to obtain plasma within 30 min after sample collection. All samples were stored at 70± 10 ◦C prior to protein precipitation and LC-MS/MS analysis (positive ion electrospray, multiple reaction monitoring mode (MRM), API 4000, AB Sciex, Framingham, MA). Standard calibration curves were con- structed by analyzing a series of control plasma aliquots containing 100 ng/mL labetol and 200 ng/mL diclofenac as internal standards (IS) and 1.0–1000 ng/mL (dogs iv) or 3000 ng/mL (rodents) EPZ015666. On all occasions, more than 75% of the non-zero standards were within 15% of deviation from the nominal concentration, including the lower limit of quantitation. Four levels of quality controls (3, 40 or 50, 800 and 2400 ng/mL) were also included in the analysis and at least 67% of the samples were within 15% of their respective nominal value, with a coefficient of variation
≤15% (at one level of quality control samples). The

concentration of EPZ015666 in each unknown sample was determined by solving the linear calibration curve equation for each corresponding drug/internal standard peak area ratio. PK parameters were calculated by noncompartmental meth- ods using Phoenix WinNonlin 6.2 (Certara, St-Louis, MO). Terminal half-life values were determined by regression of at least three data-points in the later phase of the time– concentration profile. Parameters are presented as mean ± standard deviation (SD), where applicable.

EPZ015666 stability in CD-1 mouse, SD rat, Beagle dog and HLM
Pooled HLM (mixed gender) and pooled male CD-1 mouse, SD rat and Beagle dog liver microsomes (DLM) were purchased from Corning Life Sciences (Amsterdam, the Netherlands). Microsomes (0.5 mg/mL), 0.1 M phosphate buffer pH 7.4 and 3 mM EPZ015666 (final DMSO concen- tration 0.25%) were pre-incubated at 37 ◦C prior to the addition of 1 mM NADPH to initiate the reaction. The final incubation volume was 50 mL. Negative control incubations were included where phosphate buffer was added instead of NADPH. Positive controls were diazepam and dextromethor- phan: diazepam had t1/2 values of 3.7 and 17 min in mouse and rat LM, respectively, whereas dextromethorphan t1/2 was
40 and 22 min in human and dog LM, respectively. Each compound was incubated for 0, 5, 15, 30 and 45 min. The negative controls were incubated for 45 min only. The reactions were stopped by transferring 25 mL of incubate to 50 mL methanol at the appropriate time points. The termin- ation plates were centrifuged at 2500 rpm for 20 min at 4 ◦C to precipitate the protein. Following protein precipitation, metoprolol (IS) was added to the sample supernatants prior to LC-MS/MS analysis. In vitro t½ values were determined by plotting the natural logarithm of the analyte/IS peak area ratios as a function of time, with the slope of the linear regression (—k) converted to in vitro t½ value, where t½ 0.693/k. Subsequently, intrinsic CL (CLint) was calculated as: (incubation volume/microsomal pro- tein) 0.693/t½ and scaled CL values were obtained using the well-stirred venous equilibration model (Houston, 1994; Pang & Rowland, 1977).

EPZ015666 stability in Beagle DLM in the presence of CYP inhibitors
The assay was performed as described above, in duplicate, in the presence and absence of either 1 mM quinidine (inhibitor of CYP2D) or 1 mM ketoconazole (inhibitor of CYP3A; Bogaards et al., 2000). In addition to parent depletion, formation of M1 (m/z 416.2) and M2 (m/z 402.2) was monitored by LC-MS/MS analysis. Mean percent activity remaining in the presence of inhibitors was calculated based on peak area/IS ratio. As quinidine was reported not to inhibit CYP2D activity in mouse and rat microsomes (Bogaards et al., 2000), these species were not included in the inhibition assay.
EPZ015666 stability in CD-1 mouse, SD rat, Beagle dog and human hepatocytes
Suspensions of pooled cryopreserved mixed gender human, male SD rat, male CD-1 mouse and male Beagle dog

hepatocytes were purchased from Celsis IVT (Belgium). EPZ015666 (3 mM; 0.25% DMSO final) was added to 96-well plates containing 0.25 106 cells in 0.5 mL of Williams E media to start the reaction. Plates were incubated at 37 ◦C and 50 mL aliquots were removed from the incubation mixture at 0, 5, 10, 20, 40 and 60 min and added to two volumes of methanol-containing metoprolol (internal standard) to stop the reaction. Negative control incubations with inactivated cell lysate were incubated for 60 min. Two positive controls (verapamil and umbelliferone or dextromethorphan) were included in each assay and all CL values were within historical range. All samples were centrifuged (2500 rpm, 4 ◦C, 30 min) and the supernatants were analyzed by LC-MS/ MS. In vitro t½ values were determined as described above and CLint were calculated as: (incubation volume/number of cells) 0.693/t½. Subsequently, scaled CL values were calculated using the well-stirred venous equation model (Houston, 1994; Pang & Rowland, 1977).

Metabolite identification in CD-1 mouse, SD rat, Beagle dog and human hepatocytes
EPZ015666, 10 mM, was incubated with mixed gender pooled human (10 donors), male pooled CD-1 mouse (n 29), SD rat (n 20) and Beagle dog (n 3) hepatocytes (Celsis, Baltimore, MD). Thawed cryopreserved hepatocytes (250 000 cells per incubation) were added to hepatocyte maintenance media (Lonza, Walkersville, MD), prior to the addition of EPZ015666. Reactions were incubated at 37 ◦C, in a humidified CO2 incubator, for 90 min and terminated by the addition of one volume of ice-cold acetonitrile. A negative sample without cells was also incubated for 90 min. Verapamil was used as positive control in each species and acceptance of the hepatocyte incubations was based on the qualitative formation of Phase I and Phase II metabolites of verapamil. Samples were centrifuged at 4000 rpm for 10 min, and supernatant was used for metabolite identification. Chromatography separation was performed using a Waters ACQUITY UPLC system with a binary solvent manager using 10 mM ammonium bicarbonate in water (A) and acetonitrile (B) as the mobile phase with the following gradient: gradient from 5% B to 95% B in 9 min, followed by 95% B for 1.25 min, at a flow rate of 500 mL/min. Samples were analyzed using an Acquity BEH C18 column (1.7 m,
2.1 100 mm; Waters, Milford, MA) at 60 ◦C. The injection
volume was 5 mL. A Waters Synapt G2-S mass spectrometer was used to perform MS scan ranging from 100 to 1000 Da (electrospray positive ion mode) for comprehensive analysis of metabolites. The mass spectrometer was operated with a desolvation temperature of 500 ◦C, source temperature of 120 ◦C, sample cone at 30 V, spray capillary at 1.0 kV and the MS/MS collision energy was set at 25 V. The possible chemical structures of the metabolites were deduced based on their MS1 and MS2 spectra in addition to their exact masses.

Human P450 phenotyping
Pooled HLM (n 200) were from XenoTech, LLC (Lenexa, KS). Human recombinant CYPs (rCYP1A2, 2B6, 2C9, 2C19, 2D6, 3A4 and 3A5) were provided as Supersomes®

(microsomes prepared from insect cells over-expressing cDNA for human CYPs) and obtained from Corning Life Sciences (Tewksbury, MA). In a total volume of 100 mL, 1 mM EPZ015666 was incubated with HLM or rCYPs (0.25 mg/mL) and 1 mM NADPH for 0, 30 and 120 min at 37 ◦C in 100 mM sodium phosphate buffer, pH 7.4. All reactions were stopped by the addition of 100 mL acetonitrile (ACN) containing labetalol (IS) to the incubation mixture. Samples were mixed, centrifuged to precipitate the proteins, and analyzed by LC-MS/MS. Chromatography separation was performed using an HPLC system with a binary solvent manager using 10 mM ammonium bicarbonate in water (A) and acetonitrile
(B) as the mobile phase with the following gradient: from 5% B to 70% B over 9 min, at a flow rate of 700 mL/min. The analytical column used was an XBridge BEH C18 (2.5 m,
4.6 100 mm; Waters, MA). The injection volume was 5 mL. EPZ015666, Metabolites 1 and 2, and the IS, were monitored in positive ion electrospray mode using MRM specific to each compound using an API4000 system (AB Sciex, Framingham, MA). The EPZ015666:labetalol peak area ratio measured at zero minutes was used to calculate the percent of parent remaining in subsequent time points. Furthermore, the peak area ratio of EPZ015666 at zero minutes and all metabolites were used to calculate the percent of EPZ015666 peak area represented by M1 and M2 at each time point, in a semi-quantitative manner. Quantifiable depletion of EPZ015666 in the incubations was indicated when 15% loss of parent was observed. In vitro t½ values were determined by plotting the natural logarithm of the % of EPZ015666 remaining versus time (0, 30 and 120 min), with the slope of the linear regression (—k) converted to in vitro t½ value, where t½ 0.693/k. Relative activity factor (RAF) was calculated as in Venkatakrishnan et al. (2001).

Results
PK studies in mouse, rat and dog
EPZ015666 time–concentration data following intravenous (i.v.) and oral dosing in mouse, rat and dog are shown in Figure 1. Male CD-1 mice administered a single dose of EPZ015666 at 2 mg/kg by i.v. bolus showed a moderate clearance of 30.0 ± 2.37 mL/min/kg with a volume of distri- bution at steady state (Vss) greater than total body water at
1.67 L/kg (Table 1). EPZ015666 was also moderately cleared in male Sprague-Dawley rats (36.8 ± 14.5 mL/min/kg), with a Vss of 2.32 L/kg. Mean terminal half-life (t1/2) was similar in rodents at 1.38 ± 0.445 h in mice and 1.23 ± 0.344 h in rats, and in line with mean residence time (MRT) values (0.928 ± 0.082 h in mice and 1.22 ± 0.652 h in rats). Male Beagle dogs administered 2 mg/kg EPZ015666 i.v. showed a relatively higher clearance at 46.5 ± 11.8 mL/min/kg, with a Vss of 1.97 ± 0.199 L/kg, corresponding to a short t1/2 of 0.578 ± 0.116 h and MRT of 0.727 ± 0.136 h. EPZ015666 oral bioavailability (F) was lower in dogs (39.2 ± 5.97%) than in rodent (mouse F 69.2 ± 18.1%, rat F 94.2 ± 23.3%; Table 1). There was likely some nonlinearity in the PK of EPZ015666 since F was higher than expected based on the in vitro– in vivo correlation (see below) which indicates the predom- inant clearance mechanism is hepatic metabolism.

Figure 1. The PKs of EPZ015666 determined in (a) mouse, (b) rat and (c) dog. Data are shown as concentration versus time plots of plasma concentrations (mean ± SD, n 3) following i.v. bolus (black circles, 2 mg/kg) or oral gavage (open triangles, 10 mg/kg) administration.

Table 1. PK parameters for EPZ015666 following i.v. bolus or gavage (p.o.) administration to CD-1 mice, Sprague–Dawley rats and Beagle dogs.

Mouse Rat Dog
Parameter i.v. p.o. i.v. p.o. i.v. p.o.
Dose (mg/kg) 2 10 2 10 2 10
CL (mL/min/kg) 30.0 ± 2.37 – 36.8 ± 14.5 – 46.5 ± 11.8 –
Vss (L/kg) 1.67 ± 0.210 – 2.32 ± 0.124 – 1.97 ± 0.199 –
Cmax (ng/mL) – 3500 ± 1133 – 3657 ± 375 – 1353 ± 234
tmax (h) – 0.333 ± 0.144 – 0.250 ± 0 – 0.500 ± 0
AUC0-last (h*ng/mL) 1110 ± 91.7 3847 ± 1014 1040 ± 521 4913 ± 1215 746 ± 193 1480 ± 546
AUC0-inf (h*ng/mL) 1113 ± 90.7 3850 ± 1010 1045 ± 526 4923 ± 1215 749 ± 193 1487 ± 549
t½ (h) 1.38 ± 0.445 1.62 ± 0.396 1.23 ± 0.344 1.43 ± 0.185 0.578 ± 0.116 1.04 ± 0.619
MRT (h) 0.928 ± 0.082 1.35 ± 0.181 1.22 ± 0.652 1.48 ± 0.202 0.727 ± 0.136 1.26 ± 0.148
F (%) – 69.2 ± 18.1 – 94.2 ± 23.3 – 39.2 ± 5.97
–: Not applicable.
Expressed as mean ± SD, n ¼ 3.

EPZ015666 stability in mouse, rat, dog and HLM and hepatocytes
As shown in Table 2, liver microsomal incubations supple- mented with NADPH showed scaled clearances corresponding to530% hepatic extraction (Eh) in human, mouse and rat. For these species, clearance values were higher in hepatocytes than in LM, with moderate hepatic extraction (37–54% Eh). In dogs, scaled clearances were similar in LM and hepatocytes, translating to a higher hepatic extraction (80–85% Eh). Hepatocyte clearance values were within two-fold of in vivo clearance for mouse, rat and dog indicating a robust cross- species in vitro–in vivo correlation. In LM, EPZ015666 was stable in the absence of NADPH in all species (negative control, 93% of the 0 min timepoint). Positive control compounds showed turnover consistent with historical values for all species, in both microsomal and hepatocyte incubations.

Identification of the major metabolites of EPZ015666 in mouse, rat, dog and human hepatocytes
Following a 90-min incubation at 10 mM in human and rodent hepatocytes, the parent compound EPZ015666, accounted for

more than 90% of the total peak area from LC-MS/MS analysis. In contrast, EPZ015666 represented approximately 44% of the total peak area in dog hepatocytes, in line with the shorter half-life seen in dog LM and hepatocytes, and the higher clearance observed in the dog PK study. A represen- tative extracted ion chromatogram for M1, M2, M3 and EPZ015666 in hepatocytes after a 90-min incubation is presented in Figure 2, and shows a significantly higher turnover in dog compared with rodent and human, together with increased formation of M1. M1, M2 and M3 were not present in the negative control (EPZ015666 incubated in media only). A summary of the metabolites identified is presented in Table 3.

EPZ015666
The protonated molecular ion ([M + H]+) of EPZ015666 was m/z 384.203, in agreement with the calculated exact mass of m/z 384.204. Figure 3 shows the MS1 and MS2 spectra of EPZ015666 with structurally diagnostic ions observed at m/z 123.056, 146.098, 172.114, 178.063, 207.089, 251.116 and
366.194. Metabolites showed similar fragmentation pathways,

Table 2. In vitro hepatic clearance of EPZ015666 in human, mouse, rat and dog.

Species Human Mouse Rat Dog
Scaled hepatic CL from microsomes (mL/min/kg) 55 19.7 ± 10.4 18.5 ± 2.75 26.1 ± 2.65
Hepatic extraction from microsomes (%) 525 21.9 26.4 84.5
Scaled hepatic CL from hepatocytes (mL/min/kg) 7.68 ± 2.38 48.7 ± 6.88 33.3 ± 1.86 24.9 ± 8.87
Hepatic extraction from hepatocytes (%) 37.1 54.1 47.5 80.7
Scaled clearance values are expressed as mean ± SE from the timeconcentration regression fit.

Figure 2. Representative ULPC-MS extracted ion chromatogram for M1, M2, M3 and EPZ015666 in hepatocytes after a 90 min incubation period.

which allowed the elucidation and assignment of metabolite structures.

M1
The protonated molecular ion of M1 was m/z 416.193, indicating a mass shift of +32 Da (+2O). The diagnostic ions at m/z 172.114 and 146.098 showed that no changes occurred on the tetrahydroisoquinoline moiety of EPZ015666 (Figure 4). The diagnostic fragment ions at m/z 239.080 (m/z
207 fragment from EPZ015666 + 2O), m/z 155.050 (m/z 123 + 2O) and m/z 109.040 (loss of HCO2H from m/z 155) suggest the opening of the oxetane ring and oxidation to a hydroxypropionoic acid moiety. In addition, the fragment ion at m/z 354.194 (loss of water and CO2 from m/z 416) supports the M1 structure assignment. M1 was a major metabolite in all species based on the percentage of total peak area (Table 3), and is likely formed by oxidation of the resultant aldehyde following the initial CYP-mediated oxetane ring opening.

M2
The protonated molecular ion of M2 was m/z 402.214, indicating a mass shift of +18 Da (+O + 2H). As for M1, the presence of the ions at m/z 172.114 and 146.098 showed that no changes occurred on the tetrahydroisoquinoline moiety of EPZ015666 (Figure 5). The diagnostic fragment ion at m/z 141.067 (m/z 123 + O + 2H) and m/z 225.100 (m/z 207 + O +
2H) is indicative of oxetane ring scission and oxidation. M2 is likely formed by reduction of the resultant aldehyde following the initial CYP-mediated oxetane ring opening, and repre- sented 55% of the total peak area in all species.

M3
The protonated molecular ion of M3 was m/z 328.177, indicating a mass shift of 56 Da (-C3H4O). As for M1 and M2, the presence of the ion at m/z 172.114 showed that no changes occurred on the tetrahydroisoquinoline moiety (Figure 6). In contrast, the diagnostic fragment ions at m/z
151.063 (loss of C3H4O from daughter ion m/z 207) and m/z

Table 3. Summary of major metabolites of EPZ015666 in hepatocytes.
% Total peak areaa
Mass shift (Da) m/z Selected product ions (m/z) Human Mouse Rat Dog
Parent 0 384.20 123.056, 146.098, 172.114, 178.063, 207.089, 251.116, 366.194 91 96 91 44
M1 +32 416.19 109.040, 146.098, 155.050, 172.114, 239.080, 283.107, 354.194, 398.186 55 55 55 50
M2 +18 402.21 141.067, 146.098, 172.114, 225.100, 269.126, 384.205 55 55 55 55
M3 —56 328.18 122.035, 146.098, 151.063, 172.114, 195.089, 310.168 51 51 51 55
Bold ions represent diagnostic fragment ions specific to metabolites.
aSemi-quantitative comparison since this does not take into account differences in MS ionization of the four EPZ015666-related species.

Figure 3. MS1 and MS2 spectra (a) with proposed fragmentation pathways of EPZ015666 (b).

122.035 (loss of C3H4O from daughter ion m/z 178) indicates that the oxetane moiety of EPZ015666 is likely subject to N-dealkylation. M3 was present in all species (55% of total peak area).
Other minor metabolites resulting from oxidation of the pyrimidine or the tetrahydroisoquinoline moiety were also observed ( 1.6% total peak area; data not shown). No evidence of phase II metabolism was seen in any hepatocyte incubations. The proposed metabolite scheme for EPZ015666 in hepatocytes is shown in Figure 7.

Human P450 phenotyping
No quantifiable depletion of EPZ015666 was observed following incubations in human rCYP1A2, rCYP2B6, rCYP2C9 or rCYP2C19. In contrast, depletion of EPZ015666 was observed following incubation with HLM, rCYP2D6, rCYP3A4 and rCYP3A5 (Table 4). While loss was 530% of initial parent peak area ratio in HLM, rCYP3A4 and rCYP3A5, loss in rCYP2D6 approached 95% after 120 min of incubation (t½ ¼ 28 min, CLint ¼ 9.9 mL/min/mg protein;

Figure 4. MS1 and MS2 spectra of M1 (m/z 416.193).

Figure 5. MS1 and MS2 spectrum of M2 (m/z 402.214).

Table 4). As expected since CYP2D6 represents only 2–4% of the total P450 content in human liver (Zhou, 2009), a much longer half-life of EPZ015666 was observed in HLM than that in rCYP2D6. Increasing concentrations of M1 and M2 was observed over time in the presence of HLM and rCYP2D6. M1 and M2 formation in rCYP3A4 and rCYP3A5 was

minimal, representing 0.25% of the EPZ015666 zero minute peak area ratio. As M3, and other minor metabolite formation was not monitored in this assay, it is likely that some of the parent loss observed with rCYP3A4/5 may be attributable to these metabolites. In contrast, M1 and M2 formation was extensive by rCYP2D6, representing approximately 40% and

Figure 6. MS1 and MS2 spectrum of M3 (m/z 328.177).

Figure 7. The proposed major metabolic pathways of EPZ015666 in mouse, rat, dog and human hepatocytes.

10% of the EPZ015666 zero minute peak area ratio, respectively. As t½ values were not quantifiable ( 250 min, Table 4) for CYP3A, a robust RAF approach was not feasible. Nevertheless, based on these data the CYP2D6 contribution to M1 and M2 formation in HLM was approximated to be ≤75%.

EPZ015666 stability in Beagle DLM in the presence of P450 inhibitors
As shown in Figure 8, EPZ015666 metabolism in DLM was inhibited in the presence of 1 mM quinidine (inhibitor of CYP2D) and 1 mM ketoconazole (inhibitor of CYP3A).

Table 4. Summary of reaction phenotyping data in human recombinant CYP isoforms.

Matrix
EPZ015666
(%remaining)
EPZ015666 t½
(min) M1
(% parent 0 min) M2
(% parent 0 min)
HLM 85 >250 55 55
rCYP2D6 5.5 28 41 12
rCYP3A4 83 >250 50.25 50.25
rCYP3A5 79 >250 BLQ 50.25
EPZ015666, 1 mM, was incubated with HLM or human recombinant P450s (0.5 mg/mL) for up to 120 min. No quantifiable depletion of EPZ015666 was observed with human rCYP1A2, 2B6, 2C9 or 2C19. BLQ, Below level of quantitation, set at 0.05% parent peak area at time
0. M1 and M2 formation are semi-quantitative values assuming similar MS ionization efficiency for parent and metabolites.

Figure 8. Metabolism of EPZ015666 in DLM in the presence and absence of a CYP2D inhibitor (quinidine) or CYP3A inhibitor (ketoconazole).

In parallel, M1 formation decreased in the presence of the inhibitors suggesting that both CYP2D15 and CYP3A12/26 are implicated in M1 formation. In contrast, ketoconazole had a limited impact on M2 formation, whereas a more pronounced inhibition was achieved using quinidine. The data suggest CYP2D15 is the major enzyme involved in the formation of both M1 and M2, with a lesser contribution from CYP3A in the formation of M1.

Discussion
In drug discovery, oxetanes may be used as a mimetic of alkyl moieties to favorably modulate the physicochemical properties of a lead compound (Meanwell, 2011). Although limited to favorable regioisomers, oxetanes have recently been reported as a relatively stable chemical moiety for drug discovery based on human and rodent microsomal stability studies (Burkhard et al., 2010, 2013; Stepan et al., 2011; Wuitschik et al., 2006). As metabolic stability is dependent on all the structural elements of a molecule, the impact of oxetane substitutions will vary by chemical series. In addition to increased metabolic stability, analogous oxetane derivatives of a phenylbutylamine compound also demon- strated improved solubility showing the broad impact of the polarity modulation afforded by the oxetane motif

(Wuitschik et al., 2006). In the course of our PRMT5 inhibitor lead optimization program, incorporation of a 3-substituted oxetane ring led to the discovery of EPZ015666, a compound with reduced lipophilicity and increased rodent and human microsomal stability, selected as a first-in-class PRMT5 tool compound for in vivo studies (Chan-Penebre et al., 2015). EPZ015666 exhibited a relatively low clearance in human and rodent liver microsomes, with extensive in vitro and in vivo clearance in dogs which triggered additional investigations to understand this unexpected inter-species difference. As dog is a commonly used animal model for human PK projections and non-rodent toxicology studies (Tibbitts, 2003), understanding differences in metabolism between dog, rat and the targeted patient population is key to improved preclinical extrapolation to humans. The observed species differences in clearance suggested that interspecies scaling, based on the PK profile in animals, would not be applicable to predict the metabolism and disposition of EPZ015666 in humans. Understanding the metabolic clearance of EPZ015666 in mouse was also critical as this species was used for pharmacology studies (MCL xenograft models; Chan-Penebre et al., 2015).
Metabolite identification studies in mouse, rat, dog and human hepatocytes with structural elucidation via accurate mass measurement and MS fragmentation (LC/MSn) revealed that the prominent site of metabolism of EPZ015666 was the oxetane moiety since M1 and M2 both resulted from oxidative oxetane ring opening and were observed in all species. M2 is plausibly a stable intermediate of M1 formation, generated by P450-catalyzed oxidations via an unstable hemiacetal and an aldehyde intermediate (Guengerich, 2001; Stepan et al., 2011). Additionally, the resultant aldehyde following CYP-mediated oxetane ring opening could be either reduced to the alcohol (M2) or oxidized to the carboxylic acid (M1) by a variety of ubiquitous oxidoreductases such as alcohol dehydrogenase. From a semi-quantitative analysis, M1 and M2 each repre- sented less than 5% of the total peak area in human and rodent hepatocytes. In contrast, M1 was equivalent to 50% of the total peak area in dog hepatocytes, suggesting that M1 formation rate is likely a key driver of the cross-species difference observed in EPZ015666 clearance, acknowledging that differ- ences in MS ionization do not allow for a fully quantitative comparison. An additional metabolite, M3, resulted from N-dealkylation, and the loss of the oxetane moiety. Three additional minor metabolites were present in trace amounts in some species, generated from oxidation of the pyrimidine or the tetrahydroisoquinoline moieties of EPZ015666.
In line with our discovery of EPZ015666 oxetane ring opening as a primary site of metabolism, Stepan et al. (2011) reported that oxidative metabolism and scission of the 3-substituted oxetane moiety of arylsulfonamide molecules following incubation with HLM. These authors suggested that these reactions were predominantly catalyzed by CYP3A4 due to decreased metabolite formation in HLM in the presence of ketoconazole, a selective CYP3A4 inhibitor. In the present study, we observed extensive EPZ015666 conver- sion to M1 and M2 in the presence of human recombinant CYP2D6, and to a lesser extent in the presence of rCYP3A4/5. Using these data, we estimated the relative contribution of CYP2D6 to M1 and M2 formation in HLM to be ≤75%. Observation of EPZ015666 metabolism by

CYP2D6 is in line with the general ligand-based model for this enzyme, in which a basic nitrogen is typically about 5– 7 A˚ away from the site of carbon hydroxylation (Wolff et al.,
1985). In contrast, EPZ015666 was not a substrate of human recombinant CYP1A2, CYP2B6, CYP2C9 or CYP2C19 or cytosolic aldehyde oxidase since the compound was stable in liver S9 fractions in the absence of NADPH (data not shown). Identification of EPZ015666, and potentially other basic oxetane-bearing compounds, as CYP2D6 substrates is clinic- ally relevant as this enzyme shows extensive genetic polymorphism, which can lead to significant inter-individual and ethnic differences in drug metabolism leading to variable drug exposure, efficacy and safety (Teh & Bertilsson, 2012). Importantly, the prevalence of oxetane ring opening in dog was primarily mediated by CYP2D15 and to a lesser extent by CYP3A based on chemical inhibition experiments in DLM. While CYP2D6 represents only 2–4% of the total P450 content in human liver (Zhou, 2009), CYP2D15, the dog ortholog, accounts for approximately 17–20% of the total P450 content in dog liver, which is distinctive from CYP2D members of other species (Heikkinen et al., 2015; Sakamoto et al., 1995). In contrast, CYP2D22, previously suggested to be the mouse ortholog of human CYP2D6, accounted for approximately 4% of the total cDNA from 31 constitutively expressed P450 isoforms in mouse liver (Blume et al., 2000; Choudhary et al., 2003). The difference in CYP2D expression levels across species may be an important contributing factor to the higher clearance observed for EPZ015666 in dog, versus rodent and human. Furthermore, in contrast to bufuralol 10-hydroxylation activity for which dog seems to be the most similar species to man with respect to CYP2D enzyme kinetics (Bogaards et al., 2000), M1 in vitro forma- tion rate in human was closer to rodents than dog, indicating that dog is likely not a universal model for translating the PK
of CYP2D substrates to man.
Conclusion
In summary, we observed cross-species differences in both in vitro and in vivo clearance for EPZ015666, a first-in-class orally available PRMT5 inhibitor. While the compound showed low to moderate in vitro metabolic clearance in human and rodents, extensive clearance was observed in DLM and hepatocytes, which translated to a relatively high in vivo clearance in dogs. Comparison of the metabolites of EPZ015666 allowed us to detect similarities between animal species and humans. The prominent metabolites were produced in all species, although with a higher level of formation in dog. Metabolite profiling in hepatocytes showed that M1 and M2 resulted from oxidative oxetane ring opening, reactions mediated predominantly by CYP2D isoforms in human and dog, with a minor contribution from CYP3A. This study is the first to identify an oxetane ring as a metabolic soft-spot prevalent in dog, and should improve preclinical translation to human for this class of compounds.
Acknowledgements
The authors thank the in vitro metabolism group of Quintiles, Indianapolis, for excellent technical support.

Declaration of interest
N.R., K.W.D., E.C.P., R.A.C., R.C. and N.J.W. are employees
of, and/or hold equity in, Epizyme Inc.

References
Blume N, Leonard J, Xu ZJ, et al. (2000). Characterization of Cyp2d22, a novel cytochrome P450 expressed in mouse mammary cells. Arch Biochem Biophys 381:191–204.
Bogaards JJ, Bertrand M, Jackson P, et al. (2000). Determining the best animal model for human cytochrome P450 activities: a comparison of mouse, rat, rabbit, dog, micropig, monkey and man. Xenobiotica 30: 1131–52.
Burkhard JA, Wuitschik G, Plancher JM, et al. (2013). Synthesis and stability of oxetane analogs of thalidomide and lenalidomide. Org Lett 15:4312–15.
Burkhard JA, Wuitschik G, Rogers-Evans M, et al. (2010). Oxetanes as versatile elements in drug discovery and synthesis. Angew Chem Int Ed Engl 49:9052–67.
Chan-Penebre E, Kuplast KG, Majer CR, et al. (2015). A selective inhibitor of PRMT5 with in vivo and in vitro potency in MCL models. Nat Chem Biol 11:432–7.
Choudhary D, Jansson I, Schenkman JB, et al. (2003). Comparative expression profiling of 40 mouse cytochrome P450 genes in embryonic and adult tissues. Arch Biochem Biophys 414: 91–100.
Guengerich FP. (2001). Common and uncommon cytochrome P450 reactions related to metabolism and chemical toxicity. Chem Res Toxicol 14:611–50.
Heikkinen AT, Friedlein A, Matondo M, et al. (2015). Quantitative ADME proteomics – CYP and UGT enzymes in the Beagle dog liver and intestine. Pharm Res 32:74–90.
Houston JB. (1994). Utility of in vitro drug metabolism data in predicting in vivo metabolic clearance. Biochem Pharmacol 47: 1469–79.
Meanwell NA. (2011). Synopsis of some recent tactical application of bioisosteres in drug design. J Med Chem 54:2529–91.
Pang KS, Rowland M. (1977). Hepatic clearance of drugs. I. Theoretical considerations of a ‘‘well-stirred’’ model and a ‘‘parallel tube’’ model. Influence of hepatic blood flow, plasma and blood cell binding, and the hepatocellular enzymatic activity on hepatic drug clearance. J Pharmacokinet Biopharm 5:625–53.
Sakamoto K, Kirita S, Baba T, et al. (1995). A new cytochrome P450 form belonging to the CYP2D in dog liver microsomes: purification, cDNA cloning, and enzyme characterization. Arch Biochem Biophys 319:372–82.
Scott JS, Birch AM, Brocklehurst KJ, et al. (2013). Optimisation of aqueous solubility in a series of G protein coupled receptor 119 (GPR119) agonists. Med Chem Commun 4:95–100.
Stepan AF, Karki K, McDonald WS, et al. (2011). Metabolism-directed design of oxetane-containing arylsulfonamide derivatives as μ-secre- tase Inhibitors. J Med Chem 54:7772–83.
Teh LK, Bertilsson L. (2012). Pharmacogenomics of CYP2D6: molecular genetics, interethnic differences and clinical importance. Drug Metab Pharmacokinet 27:55–67.
Tibbitts J. (2003). Issues related to the use of canines in toxicologic pathology-issues with pharmacokinetics and metabolism. Toxicol Pathol 31:17–24.
Venkatakrishnan K, von Moltke LL, Greenblatt DJ. (2001). Application of the relative activity factor approach in scaling from heterologously expressed cytochromes p450 to human liver microsomes: studies on amitriptyline as a model substrate. J Pharmacol Exp Ther 297: 326–37.
Wolff T, Distlerath LM, Worthington MT, et al. (1985). Substrate specificity of human liver Cytochrome P-450 debrisoquine 4-hydroxylase probed using immunochemical inhibition and chemical modeling. Cancer Res 45:2116–22.
Wuitschik G, Rogers-Evans M, Mu¨ller K, et al. (2006). Oxetanes as promising modules in drug discovery. Angew Chem 118:7900–3.
Zhou SF. (2009). Polymorphism of human cytochrome P450 2D6 and its clinical significance: part I. Clin Pharmacokinet 48:689–723.GSK3235025