High-Resolution Mass Spectrometry for Rapid Metabolite Characterization and Reaction Phenotyping: A Case Study with Joanna E. Barbara, Sylvie Kandel, Forrest Stanley and David B. Buckley A DIVISION OF XenoTech, LLC, 16825 W 116th St, Lenexa, KS, USA

Introduction Tables and Figures

Assessing the involvement of specific (P450) in biotransformation of a drug is an important step in evaluating overall disposition and victim potential for clinically-relevant drug-drug interactions.1 Table 1. Accurate mass metabolite profile for repaglinide (1 or 10 µM) incubated with NADPH-fortified human liver microsomes (0.1 or 0.5 mg/mL) for up to 30 min Figure 4. Effect of preincubation (30 min) of NADPH-fortified human liver microsomes (0.1 mg/mL) with the CYP3A4 inhibitor mibefradil (1 µM) and the CYP2C8 Current methods for biotransformation pathway identification (reaction phenotyping) rely on prior knowledge of inhibitor glucuronide (100 µM) on formation of five established metabolites, biotransformation routes, and availability of metabolite reference standards or radioisotopically- Retention Time Experimental Mass shift from Proposed elemental Theoretical Mass error Identity Proposed biotransformation metabolites formed from repaglinide (10 µM) incubated for 30 min labeled drug. Synthesis of metabolite standards and radiolabeled compounds is time-consuming and expensive (min) m/z repaglinide composition m/z (ppm) a and the material is typically not available in early preclinical testing. Reaction phenotyping can be achieved by M0-OH 5.48 469.2693 +15.9940 C27H36N2O5 469.2702 -1.9 Hydroxylation 100 No inhibitor monitoring loss of the parent drug, but that approach suffers from a lack of specificity. Consequently, it is C1 5.94 451.2590 -2.0156 C27H34N2O4 451.2597 -0.2 Dehydrogenation preferable to monitor the formation of one or more metabolites to obtain a comprehensive biotransformation C2 6.83 441.2369 -12.0384 C25H32N2O5 441.2389 4.7 O-Deethylation + hydroxylation 90

a ) Mibefradil M4 7.50 469.2700 +15.9947 C27H36N2O5 469.2702 -0.4 Hydroxylation 80 map for the drug of interest. In general, this requires radiolabeled drug or metabolite reference standard a % (

M1 7.95 385.2112 -68.0641 C22H28N2O4 385.2127 -3.9 N,N-Didealkylation material, potentially adding several months to the drug development timeline before biotransformation pathway l a C H N O O-Deethylation o 70 Gemfibrozil M5 8.26 425.2436 -28.0317 25 32 2 4 425.2440 -4.5 r assignments can be made. t glucuronide Repaglinide 8.98 453.2738 -0.0015 C27H36N2O4 453.2753 -3.3 Not applicable n 60 o

C3 10.09 471.2848 +18.0095 C27H38N2O5 471.2859 -0.9 Hydroxylation + reduction c f 50 High-resolution mass spectrometry (HRMS) is a powerful tool for a priori metabolite profiling and characteriza- a o M2 10.30 485.2642 +31.9889 C27H36N2O6 485.2652 -5.1 N-Dealkylation + oxidation to the carboxylic acid tion because complex data sets comprising information on all of the components in a sample, within a specified C4 16.09 451.2583 -2.0170 C27H34N2O4 451.2597 -3.1 Dehydrogenation g e 40 2 a mass range, are acquired. HRMS data can be employed for relative quantitation of all detected components C5 17.23 451.2591 -2.0162 C27H34N2O4 451.2597 -1.3 Dehydrogenation t n 30 and have potential for derivation of earlier stage reaction phenotyping information without the need for C H N O Dehydrogenation e

C6 17.52 451.2599 -2.0154 27 34 2 4 451.2597 0.4 c

a r 20 metabolite reference standards. In the present study, the prandial glucose regulator repaglinide was incubated Corresponds to a metabolite previously established in the literature e with various in vitro reaction phenotyping test systems, and the samples were analyzed by HRMS metabolite P 10 profiling approaches to characterize all of the repaglinide metabolites formed. Relative amounts of each 0 individual metabolite detected in different incubation test systems were compared to establish relationships M0-OH M1 M2 M4 M5 between specific P450 enzymes and repaglinide metabolites. In the absence of metabolite reference standards, Table 2. Comparison of cytochrome P450 enzymes identified with full scan accurate mass spectrometry and relative quantitation CYP2C8 CYP3A4 Unclea r CYP2C8 CYP3A4 the abundance values were not compared across different metabolites since ionization efficiency for each as contributing to repaglinide biotransformation with involvement established with traditional methods metabolite is unknown. The approach was assessed for its suitability for rapid determination of specific biotransformation pathways for a drug in development without the additional time and cost associated with M0-OH M1 M2 M4 M5 generating radiolabeled drug or metabolite reference standards. CYP2C8, Recombinant P450 enzyme Minor contribution CYP2C8, CYP3A4 CYP3A4 CYP2C8 CYP3A4 Figure 5. Biotransformation scheme for repaglinide established with the high-resolution mass spectrometry experiment from CYP3A4 relative quantitation approach

Correlation experiment with OH CYP2C8 CYP3A4 CYP3A4 CYP2C8 CYP3A4 CH O Materials and Methods human liver microsome panel 3 H3C O CH3

Chemical inhibition in human NH O CYP2C8 CYP3A4 CYP2C8, CYP3A4 CYP2C8 CYP3A4 liver microsomes N Repaglinide was purchased from Toronto Research Chemicals (Ontario, Canada). Human liver microsomes Overall conclusion CYP2C8 CYP3A4 CYP2C8, CYP3A4 CYP2C8 CYP3A4 Repaglinide (HLM; n=1 or n=200) were prepared and characterized in-house. Recombinant cytochrome P450 enzymes CYP2C8 3 CYP3A4 3,4 CYP2C8 3,4, CYP3A4 3,4 CYP2C8 3,4 co-expressed with reductase were purchased from Cypex (Dundee, Scotland). 3 Hydroxylation (CYP2C8) Literature Minor contribution Minor contribution Aldehyde Minor contribution CYP3A4 N-dealkylation (CYP3A4) O-deethylation (CYP3A4) 3 3 4 3 OH OH OH from CYP3A4 from CYP2C8 dehydrogenase from CYP3A4 OH

CH3 O CH3 O CH3 O Repaglinide (1 or 10 µM) was incubated for up to 30 min (37°C; pH 7.4) with NADPH-fortified HLM (0.1 or 0.5 3 CH O Bidstrup et al (2003) 3 HO H3C O CH3 H3C O CH3 H3C OH 4 mg protein/mL; pooled or from individual donors) or recombinant enzymes, namely CYP1A2, 2B6, 2C8, 2C9, Säll et al (2012) H3C O CH3 NH O NH O NH O 2C19, 2D6, 3A4, or 2J2 (10 pmol/mL). For the inhibition experiments, pooled HLM were pre-incubated with NH O N N N mibefradil (1 µM) or gemfibrozil glucuronide (100 µM), metabolism-dependent inhibitors of CYP3A4 and NH O HO CYP2C8, respectively, for 30 min before incubation with repaglinide. Following protein precipitation with Not detected M4 M0-OH M5 acetonitrile, all samples were analyzed for metabolite profiling with reverse-phase gradient ultra-performance LC and HRMS on a Waters Synapt G2 HDMS quadrupole time of flight mass spectrometer equipped with a Waters Figure 1. Representative plots showing relative abundance changes with time for M1 formed Figure 3. Variation in formation of M0-OH (a,b), M1 (c,d), M2 (e,f), Reduction (reductases/P450) N-dealkylation (CYP3A4) OH Acquity LC (Milford, MA). An unrelated internal standard (1’-hydroxymidazolam) was employed for relative from repaglinide incubated at (a) 1 µM and (b) 10 µM and M4 formed from repaglinide M4 (g,h) and M5 (i,j) with measured CYP2C8 and OH CH O incubated at (c) 1 µM and (d) 10 µM with NADPH-fortified HLM (0.1 and 0.5 mg/mL) for CYP3A4 activity. Data correspond to repaglinide (10 µM) 3 CH3 O quantitation. H C O CH up to 30 min incubated with NADPH-fortified human liver microsomes 3 3 H3C O CH3 (0.1 mg/mL) for 20 min NH O NH O NH OH Metabolites were separated on a Waters Acquity BEH column (1.7 µm, 2.1 x 100 mm) at 50°C with 0.1% formic (a) (b) NH2 Potentially C3 M1 acid in water or acetonitrile (0.4 mL/min) using the following gradient: 10% organic held for 1 min, increased to 0.014 0.14 1.00 (a) 0.90 (b) M1 0.1 mg/mL M1 0.90 0.80 0.1 mg/mL M0-OH M0-OH Oxidation (reductases/P450) 35% at 7 min, 42% at 12 min, 73% at 20 min, and 90% at 25 min. The mass spectrometer was operated in 0.012 0.12 0.80 0.70 o o i 0.5 mg/mL i OH

0.5 mg/mL t 0.70 E t

a 0.60 a r o positive, resolution elevated energy mass spectrometry (MS ) or MSMS mode with electrospray ionization. All o 0.01 0.1 0.60 r ti ti 0.50 CH3 O a E a ea

0.50 ea r r r r samples were analyzed with full scan MS . The low energy scan data were employed for quantitation. Generic a a a a 0.40 0.008 0.08 0.40 H3C O CH3 e e r r 0.30 a a 0.30 eak ionization parameters were used throughout. Fexofenadine was employed for real-time mass calibration. Data

eak NH O O k k P 0.006 0.06 0.20 P a a 0.20 e processing employed MetaboLynx XS and QuanLynx subroutines of Waters MassLynx version 4.1 (Milford, MA). e 0.10 P P r = 0.967 0.10 NH OH 0.004 0.04 0.00 r = 0.569 0.00 Metabolite structural elucidation was performed manually. 0 100 200 300 400 500 M2 0.002 0.02 0 2000 4000 6000 8000 1 µM 10 µM CYP2C8 0 0 CYP3A4/5 (pmol/mg/min) (pmol/mg/min) 0 10 20 30 0 10 20 30 Time (min) Time (min) 0.08 (c) 0.09 (d) Results 0.07 M1 0.08 M1 0.07 o o 0.06 i (c) (d) i t t 0.06 a a r 6.00 0.05 r Conclusion 0.70 0.05 ea E 0.1 mg/mL 0.04 ea r 0.1 mg/mL M4 M4 r a Eleven repaglinide metabolites were characterized by full scan accurate MS in the HLM incubations as a 0.04 0.60 5.00 0.03 3,4 0.5 mg/mL 0.5 mg/mL 0.03 eak summarized in Table 1. Five were well-established, previously described metabolites, namely M0-OH and M4 0.02 eak P P o o 0.50 0.02 • The HRMS approach was appropriate for rapid, simultaneous metabolite profiling and reaction ti ti 4.00

a 0.01 a r (hydroxylation metabolites), M1 (the N-dealkylated primary amine metabolite), M5 (O-desethyl repaglinide) and r 0.01 r = 0.927

r = 0.570 a a 0.40 0 phenotyping without the need for metabolite standards for the test compound repaglinide. e 5 e 0 r M2 (the ring-opened dicarboxylic acid and major human metabolite formed in vivo. Six other metabolites (C1 r 3.00 a a 0 100 200 300 400 500 0 2000 4000 6000 8000

k 0.30 k a a e through C6) were observed. Four (C1, C4, C5 and C6) were formed by dehydrogenation. C2 was formed from e CYP2C8 CYP3A4/5 P P 2.00 • The approach resulted in biotransformation pathway assignment for the test compound repaglinide that M5 by hydroxylation. C3 was consistent with the alcohol intermediate involved in formation of M2. 0.20 (pmol/mg/min) (pmol/mg/min) was in good agreement with published literature. 0.10 1.00 1 µM 10 µM 0.06 (e) 0.07 (f) As shown in Figure 1 (for M1 and M4), metabolite formation data (by peak area ratio) were derived for the HLM 0.00 0.00 • In early-stage drug discovery it is not always straightforward to determine which in vitro metabolites may incubation time-course to establish initial rate conditions for subsequent experiments. Formation over time was 0 10 20 30 0 10 20 30 0.05 M2 0.06 M2 o o i Time (min) Time (min) i be relevant in an in vivo situation. Therefore the availability of the additional data for subsequent data t 0.04 t 0.05 a a r linear up to 20 min for the five established metabolites at 0.1 mg/mL protein and 10 µM repaglinide. Deviations r 0.04 mining is an attractive proposition. An important benefit to this strategy is that corresponding data were ea 0.03 ea r r a from linearity were observed in the 0.5 mg/mL protein samples. Some of the metabolites (M1, M2 and M5) were a 0.03 0.02 generated for the non-established metabolites C1 through C6 (not shown). eak not detected or were present at low abundance in several of the 1 µM incubations. Subsequent experiments eak 0.02 P 0.01 P were performed with repaglinide (1 and 10 µM) incubated for up to 20 min with 0.1 mg/mL protein or 10 pmol/mL Figure 2. Extracted normalized ion chromatograms for (a) M1 (m/z 385.2127 ± 20 mDa) and (b) M4 r = 0.570 0.01 0 r = 0.719 (m/z 469.2702 ± 20 mDa) in samples of repaglinide (10 µM) incubated with 0 recombinant cytochrome P450. 0 100 200 300 400 500 NADPH-fortified recombinant cytochrome P450 enzymes (10 pmol/mL) for 20 min 0 2000 4000 6000 8000 CYP2C8 CYP3A4/5 (pmol/mg/min) Recombinant P450 enzymes which formed each of the five established metabolites are summarized in Table 2. 1A2-10 uM Repaglinide-T30 2J2-10 uM Repaglinide-T30 (pmol/mg/min) (a) (b) Acknowledgements Abundance data were normalized to the maximum observed amount of the specific metabolite of interest. With 100 100 % CYP1A2 % CYP1A2 6.00 5.00 0 0 (g) (h) the exception of C2, all of the repaglinide metabolites were also detected in the recombinant enzyme 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 4.50 5.00 M4 M4 4.00 The authors would like to acknowledge Phyllis Yerino for assistance with sample preparation. o 100 100 o

incubations. Metabolite M1 was detected at similar levels in incubations with recombinant CYP3A4 and i i t t 3.50 % % 4.00 a CYP2B6 CYP2B6 a r recombinant CYP2C8, while M4 was predominantly observed in the recombinant CYP2C8 incubations (Figure 0 0 r 3.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 ea

3.00 ea

r 2.50 r a 2). CYP2C8 formed M0-OH, M1 and M4. CYP3A4 formed M1, M2, M5 and a small amount of M0-OH. M4 a 2.00 100 M1 100 2.00 % % eak CYP2C8 CYP2C8 eak 1.50 P

Recombinant enzymes are not always representative of more complete test systems so additional experiments P 0 0 1.00 1.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 r = 0.963 were performed. 0.50 0.00 r = 0.552 100 100 0.00 % % 0 100 200 300 400 500 References CYP2C9 CYP2C9 0 2000 4000 6000 8000 0 0 Representative correlation plots for metabolite formation with CYP2C8 and CYP3A4/5 activities are shown as 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 CYP2C8 CYP3A4/5 (pmol/mg/min) Testosterone 6β-hydroxylase Figure 3. Following incubation with HLM from the individual typed donor panel, relative abundance data were 100 100 (pmol/mg/min) % CYP2C19 % CYP2C19 1. Ogilvie BW et al. (2008) Chapter 7, in: Drug-Drug Interactions (Rodrigues AD Ed), Informa Healthcare USA 0 0 derived and the peak area ratios associated with detection of each individual metabolite were plotted against the 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 Inc., New York, NY. 0.08 (i) 0.08 (j) 100 100 measured activity of specific P450 enzymes in HLM from each donor. M0-OH and M4 formation showed poor 0.07 % CYP2D6 % CYP2D6 M5 0.07 M5 0 0 o correlation with CYP3A4 activity and good correlation with CYP2C8 activity, while M1 formation correlated well o 0.06 i 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 i 0.06 t t 2. Barbara JE et al. (2013) Bioanalysis 5(10): 1211-1228. a a r M1 0.05 r 0.05 with CYP3A4 activity and poorly with CYP2C8 activity. M2 and M5 formation correlated best with CYP3A4/5 100 100 ea 0.04 ea % % r CYP3A4 CYP3A4 r 0.04 a activity. 0 0 0.03 a 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 0.03 3. Bidstrup TB et al. (2003) Br J Clin Pharmacol 56: 305-314. eak 0.02 eak P P 0.02 100 100 0.01 % P450 enzymes that showed correlation with formation of each individual established metabolite are summarized CYP2J2 % CYP2J2 r = -0.030 0.01 r = 0.745 0 Time 0 Time 0 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 6.00 6.50 7.00 7.50 8.00 8.50 9.00 9.50 10.00 0 4. Säll C et al. (2012) Drug Metab Dispos 40(7): 1279-1289. in Table 2. The recombinant enzyme and correlation data were in good agreement for M0-OH (CYP2C8), M2 0 100 200 300 400 500 0 2000 4000 6000 8000

(CYP3A4), M4 (CYP2C8) and M5 (CYP3A4), but were conflicting for M1. Therefore, an inhibition experiment CYP2C8 CYP3A4/5 (pmol/mg/min) (pmol/mg/min) employing irreversible inactivation of CYP3A4 by mibefradil and CYP2C8 by gemfibrozil glucuronide, 5. van Heiningen PNM et al. (1999) Eur J Clin Pharmacol 55: 521-525. respectively, prior to incubation of repaglinide with HLM was performed.

The inhibition results are summarized in Figure 4 and Table 2. By comparing the extent of formation of the five metabolites in HLM incubated in the presence of the inhibitors to the extent observed in the absence of inhibitor, involvement of CYP2C8 in the formation of M0-OH and M4 and of CYP3A4 in formation of M1 and M5 was established. The inhibition data were inconclusive for M2.

Table 2 summarizes the results derived with the HRMS approach (Figure 5) compared with literature results derived by traditional methods. Consistent with the literature, the approach resulted in the identification of CYP2C8 as the primary mediator of M0-OH and M4 formation and CYP3A4 as primarily involved in formation of M1 and M5. Although only CYP2C8 and 3A4 were implicated in formation of M2 in HLM, the data were ambiguous as to the extent of involvement of either enzyme. This is likely to be due to the established fact that cytosolic aldehyde dehydrogenase can play a role in M2 formation.4